Repository: dgrapov/TeachingDemos Branch: master Commit: d949dfa2435e Files: 79 Total size: 55.6 MB Directory structure: gitextract_swixnm2d/ ├── .gitignore ├── Demos/ │ ├── Biochemical and Chemical Similarity Networks/ │ │ ├── Edge list.xlsx │ │ ├── Node attributes.xlsx │ │ ├── Tutorial- Building Biochemical and Chemical Similarity Based Networks .r │ │ ├── Tutorial- Building Biochemical and Chemical Similarity Networks.pptx │ │ └── partial network.cys │ ├── Data Analysis Workflow/ │ │ ├── data/ │ │ │ └── data cube │ │ ├── data analysis workflow.R │ │ ├── images/ │ │ │ └── hsa00010.xml │ │ ├── report/ │ │ │ ├── Data Analysis Workflows 2 2015.pptx │ │ │ ├── Data Analysis Workflows 2015.pptx │ │ │ ├── Data Analysis Workflows part 1 2015.pptx │ │ │ ├── Data Analysis Workflows part 2 2015.pptx │ │ │ ├── hsa00010.xml │ │ │ ├── report.Rmd │ │ │ ├── report.html │ │ │ └── report.md │ │ └── results/ │ │ ├── IMPaLA_results.csv │ │ ├── statistical_results.csv │ │ └── statistical_results_sig.csv │ ├── Introduction to Network Mapping/ │ │ ├── Demonstration.pptx │ │ ├── biochem network edge list.xlsx │ │ ├── biochem network node attributes.xlsx │ │ ├── biochemical network.cys │ │ ├── name edge list.xlsx │ │ ├── name network.cys │ │ └── node attributes.xlsx │ ├── Mapped Network From Data (Biochemical and Structural)/ │ │ ├── .RData │ │ ├── data.csv │ │ ├── edge.list.csv │ │ ├── network code.r │ │ ├── network visualization.cys │ │ ├── node.attributes.csv │ │ ├── node.attributes.xls │ │ ├── sample info.csv │ │ ├── statistics.csv │ │ └── variable info.csv │ ├── Pathway Analysis/ │ │ ├── .xml │ │ ├── KEGG Pathway Enrichment.Rmd │ │ ├── KEGG Pathway Enrichment.html │ │ ├── KEGG Pathway Enrichment.md │ │ ├── ath00020.xml │ │ ├── ath00260.path │ │ ├── ath00260.xml │ │ ├── ath00650.xml │ │ ├── athp │ │ └── hsa00010.xml │ ├── Predictive Modeling/ │ │ ├── Iris O-PLS-DA/ │ │ │ ├── O-PLS modeling of Iris data.Rmd │ │ │ ├── O-PLS modeling of Iris data.html │ │ │ ├── O-PLS modeling of Iris data.md │ │ │ ├── O-PLS_modeling_of_Iris_data.html │ │ │ ├── O-PLS_modeling_of_Iris_data.md │ │ │ └── demo.r │ │ ├── O-PLS/ │ │ │ ├── OPLS example.Rmd │ │ │ ├── OPLS_example.html │ │ │ └── OPLS_example.md │ │ └── Single and multiple Y O-PLS modeling/ │ │ ├── OPLS example.Rmd │ │ ├── OPLS example.html │ │ ├── OPLS example.md │ │ └── OPLS example.r │ ├── R Programming/ │ │ ├── R tutorial 1.R │ │ ├── R tutorial 1.html │ │ ├── R tutorial 1.md │ │ └── R tutorial 1.rmd │ ├── Statistical Analysis/ │ │ ├── Covariate Adjustment.Rmd │ │ ├── Covariate Adjustment.html │ │ ├── Covariate Adjustment.md │ │ └── style.R │ ├── Translating Between Chemical Identifiers/ │ │ └── Translations.r │ └── dplyr/ │ ├── Tutorials.Rproj │ ├── benchmark results │ ├── data/ │ │ └── benchmark results │ ├── hands_on_with_dplyr.Rmd │ ├── hands_on_with_dplyr.html │ └── hands_on_with_dplyr.md ├── README.html ├── README.md ├── TeachingDemos.Rproj └── style.R ================================================ FILE CONTENTS ================================================ ================================================ FILE: .gitignore ================================================ # History files .Rhistory # Example code in package build process *-Ex.R .Rproj.user ================================================ FILE: Demos/Biochemical and Chemical Similarity Networks/Tutorial- Building Biochemical and Chemical Similarity Based Networks .r ================================================ #load needed functions: package devium which is stored on a github source("http://pastebin.com/raw.php?i=UyDBTA57") # get sample chemical identifiers here:https://docs.google.com/spreadsheet/ccc?key=0Ap1AEMfo-fh9dFZSSm5WSHlqMC1QdkNMWFZCeWdVbEE#gid=1 #Pubchem CIDs = cids cids # overview nrow(cids) # how many str(cids) # structure, wan't numeric cids<-as.numeric(as.character(unlist(cids))) # hack to break factor #get KEGG RPAIRS #making an edge list based on CIDs from KEGG reactant pairs KEGG.edge.list<-CID.to.KEGG.pairs(cid=cids,database=get.KEGG.pairs(),lookup=get.CID.KEGG.pairs()) head(KEGG.edge.list) dim(KEGG.edge.list) # a two column list with CID to CID connections based on KEGG RPAIS # how did I get this? #1) convert from CID to KEGG using get.CID.KEGG.pairs(), which is a table stored:https://gist.github.com/dgrapov/4964546 #2) get KEGG RPAIRS using get.KEGG.pairs() which is a table stored:https://gist.github.com/dgrapov/4964564 #3) return CID pairs #get EDGES based on chemical similarity (Tanimoto distances >0.07) tanimoto.edges<-CID.to.tanimoto(cids=cids, cut.off = .7, parallel=FALSE) head(tanimoto.edges) # how did I get this? #1) Use R package ChemmineR to querry Pubchem PUG to get molecular fingerprints #2) calculate simialrity coefficient #3) return edges with similarity above cut.off #after a little bit of formatting make combined KEGG + tanimoto edge list # https://docs.google.com/spreadsheet/ccc?key=0Ap1AEMfo-fh9dFZSSm5WSHlqMC1QdkNMWFZCeWdVbEE#gid=2 #now upload this and a sample node attribute table (https://docs.google.com/spreadsheet/ccc?key=0Ap1AEMfo-fh9dFZSSm5WSHlqMC1QdkNMWFZCeWdVbEE#gid=1) #to Cytoscape ================================================ FILE: Demos/Data Analysis Workflow/data analysis workflow.R ================================================ #------------------------ # Dmitry Grapov, PhD # CDS- Creative Data Solutions # 10/2/15 #------------------------ #------------------------ # Example of an Omics # data analysis strategies # 1) exploratory analysis # 2) statistical analysis # 3) predictive modeling # 4) functional analysis #dependencies { pkg<-c("ggplot2","dplyr","R.utils","fdrtool","caret","randomForest","pROC") lapply(pkg, function(x) { if(!require(x,character.only = TRUE)) install.packages(x,character.only = TRUE) } ) #bioConductor source("https://bioconductor.org/biocLite.R") if(!require("pcaMethods")) biocLite("pcaMethods") if(!require("pathview")) biocLite("pathview") if(!require("KEGGREST")) biocLite("KEGGREST") } #set working directory wd<-"C:/Users/Dmitry/Dropbox/Software/TeachingDemos/Demos/Data Analysis Workflow/" setwd(wd) #load devium functions { sourceDirectory( "R",recursive=TRUE,verbose=TRUE) } # load data # anonymized real-world data set # comparing sick and healthy patients # metabolite and protein profiles { load(file="data/data cube") # data.obj data.cube<-data.obj$raw #take a look at the parts str(data.cube) } # 1) exploratory analysis { #Principal Components Analysis (PCA) { #raw { args<-list( pca.data = data.cube$data, pca.algorithm = "svd", pca.components = 8, pca.center = FALSE, pca.scaling = "none", pca.cv = "q2" ) #calculate and view scree plot res<-devium.pca.calculate(args,return="list",plot=TRUE) #plot results #scores highlighting healthy and sick p.args<-list( pca = res, results = "scores", color = data.cube$sample.meta[,"group",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) #loadings p.args<-list( pca = res, results = "loadings", color = data.cube$variable.meta[,"type",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) #now try with scaling args<-list( pca.data = data.cube$data, pca.algorithm = "svd", pca.components = 8, pca.center = TRUE, pca.scaling = "uv", pca.cv = "q2" ) #calculate and view scree plot res2<-devium.pca.calculate(args,return="list",plot=TRUE) #plot results #scores highlighting healthy and sick p.args<-list( pca = res2, results = "scores", color = data.cube$sample.meta[,"group",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) #loadings p.args<-list( pca = res2, results = "loadings", color = data.cube$variable.meta[,"type",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) #scores highlighting batches p.args<-list( pca = res2, results = "scores", color = data.cube$sample.meta[,"batch",drop=FALSE], font.size =3 ) do.call("plot.PCA",p.args) } #normalized { data.cube<-data.obj$normalized #now try with scaling args<-list( pca.data = data.cube$data, pca.algorithm = "svd", pca.components = 8, pca.center = TRUE, pca.scaling = "uv", pca.cv = "q2" ) #calculate and view scree plot res<-devium.pca.calculate(args,return="list",plot=TRUE) #plot results #scores highlighting healthy and sick p.args<-list( pca = res, results = "scores", color = data.cube$sample.meta[,"group",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) #scores highlighting batches p.args<-list( pca = res, results = "scores", color = data.cube$sample.meta[,"batch",drop=FALSE], font.size =3 ) do.call("plot.PCA",p.args) } } } # 2) statistical analysis { #get summaries and t-test stats data<-data.cube$data meta<-data.cube$sample.meta[,"group",drop=FALSE] #get summary .summary<-stats.summary(data,comp.obj=meta,formula=colnames(meta),sigfigs=3,log=FALSE,rel=1,do.stats=TRUE) stats.obj<-cbind(data.cube$variable.meta,.summary) head(stats.obj) write.csv(stats.obj,file="results/statistical_results.csv") } # 3) Predictive modeling { #create a classification model using random forests #generate training/test set set.seed(998) data<-data.cube$data inTraining <- createDataPartition(data.cube$sample.meta$group, p = 2/3, list = FALSE) train.data <- data[ inTraining,] test.data <- data[-inTraining,] train.y <- data.cube$sample.meta$group[ inTraining] %>% droplevels() test.y <- data.cube$sample.meta$group[ -inTraining] %>% droplevels() #set model parameters fitControl <- trainControl(## 10-fold CV method = "repeatedcv", number = 3, ## repeated ten times repeats = 3, classProbs = TRUE, verboseIter =TRUE, summaryFunction = twoClassSummary ) #fit model to the training data set.seed(825) fit<- train(train.y ~ ., data = train.data, method = "rf", trControl = fitControl, metric = "ROC", tuneLength = 3 ) fit # create ROC curve #create ROC curve for test plot(roc(as.numeric(predict(fit,newdata=train.data)),as.numeric(train.y))) #predict the test set pred<-predict(fit,newdata=test.data) prob<-predict(fit,newdata=test.data,type="prob") obs<-test.y #get performance stats twoClassSummary(data=data.frame(obs,pred,prob),lev=levels(pred)) #create ROC curve for test plot(roc(as.numeric(pred),as.numeric(obs))) } # 4) pathway enrichment analysis { # carry out analysis to identify enriched pathways # Here used IMPaLa: http://impala.molgen.mpg.de/ # load gene names as entrez, metbolites as KEGG #format data to show fold changes in pathway #get formatted data for pathview library(KEGGREST) library(pathview) data<-stats.obj #metabolite met<-data %>% dplyr::filter(type =="metabolite") %>% dplyr::select(ID,mean.sick_mean.healthy) %>% mutate(FC=log(mean.sick_mean.healthy)) %>% dplyr::select(-mean.sick_mean.healthy) #protein prot<-data %>% dplyr::filter(type =="protein") %>% dplyr::select(ID,mean.sick_mean.healthy) %>% mutate(FC=log(mean.sick_mean.healthy)) %>% dplyr::select(-mean.sick_mean.healthy) #set rownames rownames(met)<-met[,1];met<-met[,-1,drop=FALSE] rownames(prot)<-prot[,1];prot<-prot[,-1,drop=FALSE] #select pathway to view path<-"Glycolysis / Gluconeogenesis" #get KEGG code and pathway IDs data(korg) organism <- "homo sapiens" matches <- unlist(sapply(1:ncol(korg), function(i) { agrep(organism, korg[, i]) })) (kegg.code <- korg[matches, 1, drop = F]) #get kegg pathway names library(KEGGREST) pathways <- keggList("pathway", kegg.code) #get code of our pathway of interest map<-grepl(path,pathways) %>% pathways[.] %>% names(.) %>% gsub("path:","",.) #create image setwd("./images") pv.out <- pathview(gene.data = prot, cpd.data = met, gene.idtype = "SYMBOL", pathway.id = map, species = kegg.code, out.suffix = map, keys.align = "y", kegg.native = T, match.data = T, key.pos = "topright") } ================================================ FILE: Demos/Data Analysis Workflow/images/hsa00010.xml ================================================ ================================================ FILE: Demos/Data Analysis Workflow/report/Data Analysis Workflows 2 2015.pptx ================================================ [File too large to display: 21.1 MB] ================================================ FILE: Demos/Data Analysis Workflow/report/Data Analysis Workflows 2015.pptx ================================================ [File too large to display: 21.1 MB] ================================================ FILE: Demos/Data Analysis Workflow/report/hsa00010.xml ================================================ ================================================ FILE: Demos/Data Analysis Workflow/report/report.Rmd ================================================ --- output: html_document: keep_md: yes --- #### [Part 1](http://www.slideshare.net/dgrapov/data-analysis-workflows-part-1-2015) ##### The following is an example of a data analysis strategy for an integrated metabolomic and proteomic data set. This tutorial is meant to give examples of some of the major common steps in an omic integration analysis workflow. You can check out all of the code in `/report/report.Rmd`. 1. exploratory analysis 2. statistical analysis 3. predictive modeling 4. functional analysis ```{r,warning=FALSE,echo=TRUE,message=FALSE} #set working directory wd<-"C:/Users/Dmitry/Dropbox/Software/TeachingDemos/Demos/Data Analysis Workflow/" setwd(wd) #load dependancies pkg<-c("ggplot2","dplyr","R.utils","fdrtool","caret","randomForest","pROC") out<-lapply(pkg, function(x) { if(!require(x,character.only = TRUE)) install.packages(x,character.only = TRUE) } ) #bioConductor source("https://bioconductor.org/biocLite.R") if(!require("pcaMethods")) biocLite("pcaMethods") if(!require("pathview")) biocLite("pathview") if(!require("KEGGREST")) biocLite("KEGGREST") #load devium functions #sourceDirectory( "R",recursive=TRUE) source("http://pastebin.com/raw.php?i=UyDBTA57") ``` ```{r,warning=FALSE,echo=TRUE,message=FALSE} #load data setwd(wd) load(file="data/data cube") # data.obj ``` ##### This data set contains `r ncol(data.obj$raw$data)` measurements for `r nrow(data.obj$raw$data)` samples. The samples are comprised of sick and healthy patients measured across two analytical batches. ```{r, echo=TRUE} table(data.obj$raw$sample.meta$group) table(data.obj$raw$sample.meta$batch) ``` **** ### Exploratory Analysis **** ##### A critical aspect of any data analysis should be to carry out an exploratory data analysis to see if there are any strange trends. Below is an example of a Principal Components Analysis (PCA). Lets start by looking at the raw data and caclculate PCA with out anys scaling. ##### PCA has three main components we can use to evaluate our data. ##### 1. Variance explained by each component ```{r, echo=TRUE,message=FALSE,warning=FALSE} data.cube<-data.obj$raw args<-list( pca.data = data.cube$data, pca.algorithm = "svd", pca.components = 8, pca.center = FALSE, pca.scaling = "none", pca.cv = "q2" ) #calculate and view scree plot res<-devium.pca.calculate(args,return="list",plot=TRUE) ``` ##### The scree plot above shows the total variance in the data explained (top) and the cumulative varince explained (bottom) by each principal component (PC). The green bars in the bottom plot show the cross-validated variance explained which can be used to give us an idea bout the stability of calculated components. How many PCs to keep can be determined based on a few criteria 1) each PC should explain some minnimum variance and 2) calculate enough PCS to explain some target variance. The hashed line in the top plot shows PCs which explain less than 1% variance and the hashed line in the bottom plot shows how many PCs arerequired to explain 80% of the varince in the data. Based on an evaluation of the scree plot we may select 2 or 3 PCs. The cross-validated varince explained (green bars) also suggest that the variance explained does not increase after the first 2 PCs. ##### 2. The sample scores can be used to visualize multivariete similarities in samples given all the varibles for each PC. Lets plot the scores and highlight the sick and healthy groups. ```{r, echo=TRUE,message=FALSE,warning=FALSE} #plot results #scores highlighting healthy and sick p.args<-list( pca = res, results = "scores", color = data.cube$sample.meta[,"group",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) ``` #### Based on the scores above the sick and healthy samples look fairly similiar. Lets next look at the variable loadings. #### 3. Variable loadings show the contribution of each varible to the calculated scores. ```{r, echo=TRUE,message=FALSE,warning=FALSE} #loadings p.args<-list( pca = res, results = "loadings", color = data.cube$variable.meta[,"type",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) ``` #### Evaluation of the loadings suggest that variance variables X838 abd X454 explain ~90% of the varince in the data. Because we did not scale the data before conducting PCA, variables with the largest magnitude will contribute most to varince explained. #### Next lets recalculate the PCA and mean center and scale all the variables by their standard deviation (autoscale). #### Variance explained ```{r, echo=TRUE,message=FALSE,warning=FALSE} args<-list( pca.data = data.cube$data, pca.algorithm = "svd", pca.components = 8, pca.center = TRUE, pca.scaling = "uv", pca.cv = "q2" ) #calculate and view scree plot res2<-devium.pca.calculate(args,return="list",plot=TRUE) ``` #### Variable loadings ```{r, echo=TRUE,message=FALSE,warning=FALSE} #loadings p.args<-list( pca = res2, results = "loadings", color = data.cube$variable.meta[,"type",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) ``` #### Sample scores ```{r, echo=TRUE,message=FALSE,warning=FALSE} #plot results #scores highlighting healthy and sick p.args<-list( pca = res2, results = "scores", color = data.cube$sample.meta[,"group",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) ``` #### There are some noticible differences in PCA after we scaled our data. 1. Variable magnitude no longer drives the majority of the variance. 2. We can see more resolution in variable loadings for the first 2 PCs. 3. There is an unexplained group structure in the score. #### Next we can try mapping other meta data to score to see if we can explain the cluster pattern. Lets show the analytical batches in the samples scores. ```{r, echo=TRUE,message=FALSE,warning=FALSE} p.args<-list( pca = res2, results = "scores", color = data.cube$sample.meta[,"batch",drop=FALSE], font.size =3 ) do.call("plot.PCA",p.args) ``` #### We can see in the scores above that the analytical batch nicely explains 35% of the varince in the data. This is a common problem in large data sets which is best handled using various data normalization methods. Here is some more information about implementing data normalizations. ###### [Metabolomics and Beyond: Challenges and Strategies for Next-generation Omic Analyses](https://imdevsoftware.files.wordpress.com/2015/09/clipboard01.png?w=300&h=225) [![Metabolomics and Beyond: Challenges and Strategies for Next-generation Omic Analyses](https://imdevsoftware.files.wordpress.com/2015/09/clipboard01.png?w=300&h=225)](https://www.youtube.com/watch?v=4AhBN5Q1oMs) ##### [Evaluation of data normalization methods](http://www.slideshare.net/dgrapov/case-study-metabolomic-data-normalization-example) **** #### [Part 3](hhttp://www.slideshare.net/dgrapov/data-analysis-workflows-part-2-2015?related=1) ### Statistical Analysis **** ##### Next lets carry out a statistical analysis and summarise the changes between the sick and ghealthy groups. Below we identify significantly altered analytes using a basic t-test with adjustment for multiple hypotheses tested. We probably want to use more sophisticated and non-parametric tests for real applications. ```{r, echo=TRUE,message=FALSE,warning=FALSE} #get summaries and t-test stats data.cube<-data.obj$normalized data<-data.cube$data meta<-data.cube$sample.meta[,"group",drop=FALSE] #get summary .summary<-stats.summary(data,comp.obj=meta,formula=colnames(meta),sigfigs=3,log=FALSE,rel=1,do.stats=TRUE) stats.obj<-cbind(data.cube$variable.meta,.summary) stats.obj %>% arrange(group_p.values) %>% head(.) #write.csv(stats.obj,file="results/statistical_results.csv") ``` #### We can visualize the differences in means for the top most altered metabolite and protein as a box plot. ```{r, echo=TRUE,message=FALSE,warning=FALSE} top.met<-stats.obj %>% filter(type =="metabolite") %>% arrange(group_p.values) %>% dplyr::select(ID) %>% dplyr:: slice(.,1) %>% unlist(.) %>% as.character(.) id<-as.character(stats.obj$ID) %in% top.met tmp.data<-data.frame(data[,id,drop=FALSE],meta) #make plot ggplot(tmp.data,aes_string(x="group",y=colnames(tmp.data)[1],fill="group")) + geom_boxplot() + ggtitle(as.character(stats.obj$description[id])) +ylab("") top.met<-stats.obj %>% filter(type =="protein") %>% arrange(group_p.values) %>% dplyr::select(ID) %>% dplyr:: slice(.,1) %>% unlist(.) %>% as.character(.) id<-as.character(stats.obj$ID) %in% top.met tmp.data<-data.frame(data[,id,drop=FALSE],meta) #make plot ggplot(tmp.data,aes_string(x="group",y=colnames(tmp.data)[1],fill="group")) + geom_boxplot() + ggtitle(as.character(stats.obj$description[id])) +ylab("") ``` **** ### Predictive Modeling **** #### Next we can try a generate a non-linear multivarite classification model to identify important variables in our data. Below we will train and validate a random forest classifier. The full data set is split into 2/3 trainning and 1/3 test set while keeping the propotion of sick and healthy samples equivalent. The model is trained using 3-fold cross-validation repeated 3 times and the ```mtry``` parameter is optimized to maximize the are under the reciever operator characteristic curve (AUCROC). ```{r, echo=TRUE,message=FALSE,warning=FALSE} #create a classification model using random forests #generate training/test set set.seed(998) data<-data.cube$data inTraining <- createDataPartition(data.cube$sample.meta$group, p = 2/3, list = FALSE) train.data <- data[ inTraining,] test.data <- data[-inTraining,] train.y <- data.cube$sample.meta$group[ inTraining] %>% droplevels() test.y <- data.cube$sample.meta$group[ -inTraining] %>% droplevels() #set model parameters fitControl <- trainControl(## 10-fold CV method = "repeatedcv", number = 3, ## repeated ten times repeats = 3, classProbs = TRUE, summaryFunction = twoClassSummary ) #fit model to the training data set.seed(825) fit<- train(train.y ~ ., data = train.data, method = "rf", trControl = fitControl, metric = "ROC", tuneLength = 3 ) ``` #### Below the optimal model is chosen while varying the ```mtry``` or the number of variables randomly sampled as candidates at each split. ```{r, echo=TRUE,message=FALSE,warning=FALSE} fit ```` #### Next we can evaluate the model performance based on predictions for the test set. We can also look at the ROC curve. ```{r, echo=TRUE,message=FALSE,warning=FALSE} #predict the test set pred<-predict(fit,newdata=test.data) prob<-predict(fit,newdata=test.data,type="prob") obs<-test.y table(pred,obs) #get performance stats twoClassSummary(data=data.frame(obs,pred,prob),lev=levels(pred)) ``` #### We can also look at the ROC curve. ```{r, echo=TRUE,message=FALSE,warning=FALSE} x<-roc(obs,prob[,levels(pred)[1]],silent = TRUE) plot(x) ``` #### Having validated our model next we can look at the most important variables driving the classification. We can look at the differences in performance when each variable is randomly permuted or the VIP. ```{r, echo=TRUE,message=FALSE,warning=FALSE} #need to get variable names vip<-varImp(fit)$importance # need to keep rownames vip<-vip[order(vip[,1],decreasing=TRUE),,drop=FALSE][1:10,,drop=FALSE] id<-colnames(train.data) %in% gsub('`','',rownames(vip)) tmp.data<-data.frame(importance=vip[,1],variable=factor(stats.obj$description[id],levels=stats.obj$description[id],ordered=FALSE)) #plot ggplot(tmp.data, aes(x=variable,y=importance)) + geom_bar(stat="identity") + coord_flip() id<-as.character(stats.obj$description) %in% as.character(tmp.data[1,2]) tmp.data<-data.frame(data[,id,drop=FALSE],meta) #make plot ggplot(tmp.data,aes_string(x="group",y=colnames(tmp.data)[1],fill="group")) + geom_boxplot() + ggtitle(as.character(stats.obj$description[id])) +ylab("") ``` **** ### Functional Analysis **** #### Finally we can identify enriched biological pathways based on the integrated changes in genes and proteins. [IMPaLA: Integrated Molecular Pathway Level Analysis](http://impala.molgen.mpg.de/) can be used to calculate enriched pathways in genes or proteins and metabolites.To do this we can querry the significantly alterd proteins and metabolites for enriched pathways (see `results/statistical_results_sig.csv`). We can view the full analysis results in `results/IMPaLA_results.csv`. next lets take an enriched pathway and fisualize the fold changes between sick and healthy in the enriched species. ```{r, echo=TRUE,message=FALSE,warning=FALSE} #format data to show fold changes in pathway #get formatted data for pathview library(KEGGREST) library(pathview) data<-stats.obj #metabolite met<-data %>% dplyr::filter(type =="metabolite") %>% dplyr::select(ID,mean.sick_mean.healthy) %>% mutate(FC=log(mean.sick_mean.healthy)) %>% dplyr::select(-mean.sick_mean.healthy) #protein prot<-data %>% dplyr::filter(type =="protein") %>% dplyr::select(ID,mean.sick_mean.healthy) %>% mutate(FC=log(mean.sick_mean.healthy)) %>% dplyr::select(-mean.sick_mean.healthy) #set rownames rownames(met)<-met[,1];met<-met[,-1,drop=FALSE] rownames(prot)<-prot[,1];prot<-prot[,-1,drop=FALSE] #select pathway to view path<-"Glycolysis / Gluconeogenesis" ``` #### Lets take a looka at the `r path` pathway. Our data needs to be formatted as below. You can also take a look at the following more detailed example of [mapping fold changes to biochemical pathways](https://github.com/dgrapov/TeachingDemos/blob/master/Demos/Pathway%20Analysis/KEGG%20Pathway%20Enrichment.md). #### Metabolite data showing KEGG ids and log fold change ```{r, echo=TRUE,message=FALSE,warning=FALSE} head(met) ``` #### Protein data showing the Entrez gene name and log fold changes ```{r, echo=TRUE,message=FALSE,warning=FALSE} head(prot) ``` #### Next we need to get the pathway code for or pathway of interest. ```{r, echo=TRUE,message=FALSE,warning=FALSE} data(korg) organism <- "homo sapiens" matches <- unlist(sapply(1:ncol(korg), function(i) { agrep(organism, korg[, i]) })) (kegg.code <- korg[matches, 1, drop = F]) ``` #### Now we can visualize the changes between sick and healthy in the `r path` pathway. ```{r, echo=TRUE,message=FALSE,warning=FALSE} setwd(wd) pathways <- keggList("pathway", kegg.code) #get code of our pathway of interest map<-grepl(path,pathways) %>% pathways[.] %>% names(.) %>% gsub("path:","",.) map #create image setwd("report") pv.out <- pathview(gene.data = prot, cpd.data = met, gene.idtype = "SYMBOL", pathway.id = map, species = kegg.code, out.suffix = map, keys.align = "y", kegg.native = T, match.data = T, key.pos = "topright") ``` ![](hsa00010.hsa00010.png) **** #### This concludes this short tutorial. You may also find the following links useful. * [Software tools](https://github.com/dgrapov) * [More examples and demos](https://imdevsoftware.wordpress.com/) © Dmitry Grapov (2015) Creative Commons License ================================================ FILE: Demos/Data Analysis Workflow/report/report.html ================================================

Part 1

The following is an example of a data analysis strategy for an integrated metabolomic and proteomic data set. This tutorial is meant to give examples of some of the major common steps in an omic integration analysis workflow. You can check out all of the code in /report/report.Rmd.

  1. exploratory analysis

  2. statistical analysis

  3. predictive modeling

  4. functional analysis

#set working directory
wd<-"C:/Users/Dmitry/Dropbox/Software/TeachingDemos/Demos/Data Analysis Workflow/"
setwd(wd)

#load dependancies
pkg<-c("ggplot2","dplyr","R.utils","fdrtool","caret","randomForest","pROC")
out<-lapply(pkg, function(x) {
  if(!require(x,character.only = TRUE)) install.packages(x,character.only = TRUE)
  }
)


#bioConductor
source("https://bioconductor.org/biocLite.R")
if(!require("pcaMethods")) biocLite("pcaMethods")
if(!require("pathview")) biocLite("pathview")
if(!require("KEGGREST")) biocLite("KEGGREST")

#load devium functions
#sourceDirectory( "R",recursive=TRUE)
source("http://pastebin.com/raw.php?i=UyDBTA57")
#load data
setwd(wd)
load(file="data/data cube") # data.obj
This data set contains 200 measurements for 54 samples. The samples are comprised of sick and healthy patients measured across two analytical batches.
table(data.obj$raw$sample.meta$group)
## 
## healthy    sick 
##      27      27
table(data.obj$raw$sample.meta$batch)
## 
## batch_1 batch_2 
##      28      26

Exploratory Analysis

A critical aspect of any data analysis should be to carry out an exploratory data analysis to see if there are any strange trends. Below is an example of a Principal Components Analysis (PCA). Lets start by looking at the raw data and caclculate PCA with out anys scaling.
PCA has three main components we can use to evaluate our data.
1. Variance explained by each component
data.cube<-data.obj$raw
args<-list(  pca.data       = data.cube$data,
            pca.algorithm   = "svd",
            pca.components  = 8,
            pca.center      = FALSE,
            pca.scaling     = "none",
            pca.cv          = "q2"
            )
            
#calculate and view scree plot          
res<-devium.pca.calculate(args,return="list",plot=TRUE)

The scree plot above shows the total variance in the data explained (top) and the cumulative varince explained (bottom) by each principal component (PC). The green bars in the bottom plot show the cross-validated variance explained which can be used to give us an idea bout the stability of calculated components. How many PCs to keep can be determined based on a few criteria 1) each PC should explain some minnimum variance and 2) calculate enough PCS to explain some target variance. The hashed line in the top plot shows PCs which explain less than 1% variance and the hashed line in the bottom plot shows how many PCs arerequired to explain 80% of the varince in the data. Based on an evaluation of the scree plot we may select 2 or 3 PCs. The cross-validated varince explained (green bars) also suggest that the variance explained does not increase after the first 2 PCs.
2. The sample scores can be used to visualize multivariete similarities in samples given all the varibles for each PC. Lets plot the scores and highlight the sick and healthy groups.
#plot results
#scores highlighting healthy and sick
p.args<-list( 
        pca = res,
            results = "scores",
            color = data.cube$sample.meta[,"group",drop=FALSE],
            font.size = 3
            )

do.call("plot.PCA",p.args)

Based on the scores above the sick and healthy samples look fairly similiar. Lets next look at the variable loadings.

3. Variable loadings show the contribution of each varible to the calculated scores.

#loadings
p.args<-list( 
        pca = res,
            results = "loadings",
            color = data.cube$variable.meta[,"type",drop=FALSE],
            font.size = 3
            )

do.call("plot.PCA",p.args)

Evaluation of the loadings suggest that variance variables X838 abd X454 explain ~90% of the varince in the data. Because we did not scale the data before conducting PCA, variables with the largest magnitude will contribute most to varince explained.

Next lets recalculate the PCA and mean center and scale all the variables by their standard deviation (autoscale).

Variance explained

args<-list(  pca.data       = data.cube$data,
            pca.algorithm   = "svd",
            pca.components  = 8,
            pca.center      = TRUE,
            pca.scaling     = "uv",
            pca.cv          = "q2"
            )
            
#calculate and view scree plot          
res2<-devium.pca.calculate(args,return="list",plot=TRUE)

Variable loadings

#loadings
p.args<-list( 
            pca = res2,
            results = "loadings",
            color = data.cube$variable.meta[,"type",drop=FALSE],
            font.size = 3
            )

do.call("plot.PCA",p.args)

Sample scores

#plot results
#scores highlighting healthy and sick
p.args<-list( 
            pca = res2,
            results = "scores",
            color = data.cube$sample.meta[,"group",drop=FALSE],
            font.size = 3
            )

do.call("plot.PCA",p.args)

There are some noticible differences in PCA after we scaled our data.

  1. Variable magnitude no longer drives the majority of the variance.
  2. We can see more resolution in variable loadings for the first 2 PCs.
  3. There is an unexplained group structure in the score.

Next we can try mapping other meta data to score to see if we can explain the cluster pattern. Lets show the analytical batches in the samples scores.

p.args<-list( 
        pca = res2,
            results = "scores",
            color = data.cube$sample.meta[,"batch",drop=FALSE],
            font.size =3
            )

do.call("plot.PCA",p.args)

We can see in the scores above that the analytical batch nicely explains 35% of the varince in the data. This is a common problem in large data sets which is best handled using various data normalization methods. Here is some more information about implementing data normalizations.

Statistical Analysis

Next lets carry out a statistical analysis and summarise the changes between the sick and ghealthy groups. Below we identify significantly altered analytes using a basic t-test with adjustment for multiple hypotheses tested. We probably want to use more sophisticated and non-parametric tests for real applications.
#get summaries and t-test stats
data.cube<-data.obj$normalized
data<-data.cube$data
meta<-data.cube$sample.meta[,"group",drop=FALSE] 

#get summary
.summary<-stats.summary(data,comp.obj=meta,formula=colnames(meta),sigfigs=3,log=FALSE,rel=1,do.stats=TRUE)
## Generating data summary... 
## Conducting tests... 
## Conducting FDR corrections...
stats.obj<-cbind(data.cube$variable.meta,.summary)
stats.obj %>% arrange(group_p.values) %>% head(.)
##       ID                     description       type
## 1 C00077                       ornithine metabolite
## 2 C02477                tocopherol alpha metabolite
## 3 C00097                        cysteine metabolite
## 4 C00031                         glucose metabolite
## 5 C00170 5'-deoxy-5'-methylthioadenosine metabolite
## 6 C00385                        xanthine metabolite
##   healthy.mean.....std.dev sick.mean.....std.dev mean.sick_mean.healthy
## 1            7300 +/- 2900         3910 +/- 2100                   0.54
## 2            1160 +/- 1400         2600 +/- 1200                   2.24
## 3            5160 +/- 2500         8460 +/- 3400                   1.64
## 4         335000 +/- 2e+05     144000 +/- 140000                   0.43
## 5               216 +/- 83           355 +/- 160                   1.64
## 6              574 +/- 350         1670 +/- 1500                   2.91
##   group_p.values group_adjusted.p.values group_q.values
## 1   9.433317e-06             0.001886663    0.000900412
## 2   1.505034e-04             0.007714558    0.003632462
## 3   1.507843e-04             0.007714558    0.003633836
## 4   1.542912e-04             0.007714558    0.003650646
## 5   2.004055e-04             0.008016221    0.003825749
## 6   3.909048e-04             0.013030159    0.006072846
#write.csv(stats.obj,file="results/statistical_results.csv")

We can visualize the differences in means for the top most altered metabolite and protein as a box plot.

top.met<-stats.obj %>% filter(type =="metabolite") %>% 
  arrange(group_p.values) %>% dplyr::select(ID) %>% 
  dplyr:: slice(.,1) %>% unlist(.) %>% as.character(.)
id<-as.character(stats.obj$ID) %in% top.met
tmp.data<-data.frame(data[,id,drop=FALSE],meta)
#make plot
ggplot(tmp.data,aes_string(x="group",y=colnames(tmp.data)[1],fill="group")) + 
  geom_boxplot() + ggtitle(as.character(stats.obj$description[id])) +ylab("")

top.met<-stats.obj %>% filter(type =="protein") %>% 
  arrange(group_p.values) %>% dplyr::select(ID) %>% 
  dplyr:: slice(.,1) %>% unlist(.) %>% as.character(.)
id<-as.character(stats.obj$ID) %in% top.met
tmp.data<-data.frame(data[,id,drop=FALSE],meta)
#make plot
ggplot(tmp.data,aes_string(x="group",y=colnames(tmp.data)[1],fill="group")) + 
  geom_boxplot() + ggtitle(as.character(stats.obj$description[id])) +ylab("")

Predictive Modeling

Next we can try a generate a non-linear multivarite classification model to identify important variables in our data. Below we will train and validate a random forest classifier. The full data set is split into 2/3 trainning and 1/3 test set while keeping the propotion of sick and healthy samples equivalent. The model is trained using 3-fold cross-validation repeated 3 times and the mtry parameter is optimized to maximize the are under the reciever operator characteristic curve (AUCROC).

#create a classification model using random forests
#generate training/test set
set.seed(998)
data<-data.cube$data
inTraining <- createDataPartition(data.cube$sample.meta$group, p = 2/3, list = FALSE)
train.data <- data[ inTraining,]
test.data  <- data[-inTraining,]
train.y <- data.cube$sample.meta$group[ inTraining] %>% droplevels()
test.y <- data.cube$sample.meta$group[ -inTraining] %>% droplevels()

#set model parameters
fitControl <- trainControl(## 10-fold CV
                          method = "repeatedcv",
                          number = 3,
                          ## repeated ten times
                          repeats = 3,
                          classProbs = TRUE,
                          summaryFunction = twoClassSummary
                )
                           
                           
#fit model to the training data
set.seed(825)
fit<- train(train.y ~ ., data = train.data,
            method = "rf",
            trControl = fitControl,
            metric = "ROC",
            tuneLength = 3  
         )

Below the optimal model is chosen while varying the mtry or the number of variables randomly sampled as candidates at each split.

fit
## Random Forest 
## 
##  36 samples
## 199 predictors
##   2 classes: 'healthy', 'sick' 
## 
## No pre-processing
## Resampling: Cross-Validated (3 fold, repeated 3 times) 
## 
## Summary of sample sizes: 24, 24, 24, 24, 24, 24, ... 
## 
## Resampling results across tuning parameters:
## 
##   mtry  ROC        Sens       Spec       ROC SD      Sens SD    Spec SD  
##     2   0.7870370  0.7222222  0.7777778  0.08098544  0.2041241  0.1666667
##   101   0.8549383  0.7777778  0.7592593  0.10090044  0.1443376  0.2060055
##   200   0.8750000  0.8333333  0.7222222  0.09107554  0.1178511  0.2204793
## 
## ROC was used to select the optimal model using  the largest value.
## The final value used for the model was mtry = 200.

Next we can evaluate the model performance based on predictions for the test set. We can also look at the ROC curve.

#predict the test set
pred<-predict(fit,newdata=test.data)
prob<-predict(fit,newdata=test.data,type="prob")
obs<-test.y
table(pred,obs)
##          obs
## pred      healthy sick
##   healthy       7    2
##   sick          2    7
#get performance stats
twoClassSummary(data=data.frame(obs,pred,prob),lev=levels(pred))
##       ROC      Sens      Spec 
## 0.9135802 0.7777778 0.7777778

We can also look at the ROC curve.

x<-roc(obs,prob[,levels(pred)[1]],silent = TRUE)
plot(x)

## 
## Call:
## roc.default(response = obs, predictor = prob[, levels(pred)[1]],     silent = TRUE)
## 
## Data: prob[, levels(pred)[1]] in 9 controls (obs healthy) > 9 cases (obs sick).
## Area under the curve: 0.9136

Having validated our model next we can look at the most important variables driving the classification. We can look at the differences in performance when each variable is randomly permuted or the VIP.

#need to get variable names
vip<-varImp(fit)$importance # need to keep rownames
vip<-vip[order(vip[,1],decreasing=TRUE),,drop=FALSE][1:10,,drop=FALSE]
id<-colnames(train.data) %in% gsub('`','',rownames(vip))
tmp.data<-data.frame(importance=vip[,1],variable=factor(stats.obj$description[id],levels=stats.obj$description[id],ordered=FALSE))

#plot
ggplot(tmp.data, aes(x=variable,y=importance)) + geom_bar(stat="identity") + coord_flip()

id<-as.character(stats.obj$description) %in% as.character(tmp.data[1,2])
tmp.data<-data.frame(data[,id,drop=FALSE],meta)
#make plot
ggplot(tmp.data,aes_string(x="group",y=colnames(tmp.data)[1],fill="group")) + 
  geom_boxplot() + ggtitle(as.character(stats.obj$description[id])) +ylab("")

Functional Analysis

Finally we can identify enriched biological pathways based on the integrated changes in genes and proteins. IMPaLA: Integrated Molecular Pathway Level Analysis can be used to calculate enriched pathways in genes or proteins and metabolites.To do this we can querry the significantly alterd proteins and metabolites for enriched pathways (see results/statistical_results_sig.csv). We can view the full analysis results in results/IMPaLA_results.csv. next lets take an enriched pathway and fisualize the fold changes between sick and healthy in the enriched species.

#format data to show fold changes in pathway
#get formatted data for pathview
library(KEGGREST)
library(pathview)
data<-stats.obj
#metabolite
met<-data %>% dplyr::filter(type =="metabolite") %>%
  dplyr::select(ID,mean.sick_mean.healthy) %>% 
    mutate(FC=log(mean.sick_mean.healthy)) %>% dplyr::select(-mean.sick_mean.healthy)
#protein
prot<-data %>% dplyr::filter(type =="protein") %>%
    dplyr::select(ID,mean.sick_mean.healthy) %>% 
    mutate(FC=log(mean.sick_mean.healthy)) %>% dplyr::select(-mean.sick_mean.healthy)

#set rownames
rownames(met)<-met[,1];met<-met[,-1,drop=FALSE]
rownames(prot)<-prot[,1];prot<-prot[,-1,drop=FALSE] 

#select pathway to view
path<-"Glycolysis / Gluconeogenesis"

Lets take a looka at the Glycolysis / Gluconeogenesis pathway. Our data needs to be formatted as below. You can also take a look at the following more detailed example of mapping fold changes to biochemical pathways.

Metabolite data showing KEGG ids and log fold change

head(met)
##                  FC
## C00379   0.41871033
## C00385   1.06815308
## C00105   0.07696104
## C00299  -0.24846136
## C00366   0.33647224
## C00086  -0.05129329

Protein data showing the Entrez gene name and log fold changes

head(prot)
##                 FC
## SPTAN1  -0.3424903
## CFH      0.1133287
## VPS13C   0.3148107
## XRCC6    1.0715836
## APOA1   -0.1392621
## SUPT16H  0.7129498

Next we need to get the pathway code for or pathway of interest.

data(korg)
organism <- "homo sapiens"
matches <- unlist(sapply(1:ncol(korg), function(i) {
    agrep(organism, korg[, i])
}))
(kegg.code <- korg[matches, 1, drop = F])
##      kegg.code
## [1,] "hsa"

Now we can visualize the changes between sick and healthy in the Glycolysis / Gluconeogenesis pathway.

setwd(wd)
pathways <- keggList("pathway", kegg.code)
#get code of our pathway of interest
map<-grepl(path,pathways) %>% pathways[.] %>% names(.) %>% gsub("path:","",.)
map
## [1] "hsa00010"
#create image
setwd("report")
pv.out <- pathview(gene.data = prot, cpd.data = met, gene.idtype = "SYMBOL", 
    pathway.id = map, species = kegg.code, out.suffix = map, keys.align = "y", 
    kegg.native = T, match.data = T, key.pos = "topright")

================================================ FILE: Demos/Data Analysis Workflow/report/report.md ================================================ #### [Part 1](http://www.slideshare.net/dgrapov/data-analysis-workflows-part-1-2015) ##### The following is an example of a data analysis strategy for an integrated metabolomic and proteomic data set. This tutorial is meant to give examples of some of the major common steps in an omic integration analysis workflow. You can check out all of the code in `/report/report.Rmd`. 1. exploratory analysis 2. statistical analysis 3. predictive modeling 4. functional analysis ```r #set working directory wd<-"C:/Users/Dmitry/Dropbox/Software/TeachingDemos/Demos/Data Analysis Workflow/" setwd(wd) #load dependancies pkg<-c("ggplot2","dplyr","R.utils","fdrtool","caret","randomForest","pROC") out<-lapply(pkg, function(x) { if(!require(x,character.only = TRUE)) install.packages(x,character.only = TRUE) } ) #bioConductor source("https://bioconductor.org/biocLite.R") if(!require("pcaMethods")) biocLite("pcaMethods") if(!require("pathview")) biocLite("pathview") if(!require("KEGGREST")) biocLite("KEGGREST") #load devium functions #sourceDirectory( "R",recursive=TRUE) source("http://pastebin.com/raw.php?i=UyDBTA57") ``` ```r #load data setwd(wd) load(file="data/data cube") # data.obj ``` ##### This data set contains 200 measurements for 54 samples. The samples are comprised of sick and healthy patients measured across two analytical batches. ```r table(data.obj$raw$sample.meta$group) ``` ``` ## ## healthy sick ## 27 27 ``` ```r table(data.obj$raw$sample.meta$batch) ``` ``` ## ## batch_1 batch_2 ## 28 26 ``` **** ### Exploratory Analysis **** ##### A critical aspect of any data analysis should be to carry out an exploratory data analysis to see if there are any strange trends. Below is an example of a Principal Components Analysis (PCA). Lets start by looking at the raw data and caclculate PCA with out anys scaling. ##### PCA has three main components we can use to evaluate our data. ##### 1. Variance explained by each component ```r data.cube<-data.obj$raw args<-list( pca.data = data.cube$data, pca.algorithm = "svd", pca.components = 8, pca.center = FALSE, pca.scaling = "none", pca.cv = "q2" ) #calculate and view scree plot res<-devium.pca.calculate(args,return="list",plot=TRUE) ``` ![](report_files/figure-html/unnamed-chunk-4-1.png) ##### The scree plot above shows the total variance in the data explained (top) and the cumulative varince explained (bottom) by each principal component (PC). The green bars in the bottom plot show the cross-validated variance explained which can be used to give us an idea bout the stability of calculated components. How many PCs to keep can be determined based on a few criteria 1) each PC should explain some minnimum variance and 2) calculate enough PCS to explain some target variance. The hashed line in the top plot shows PCs which explain less than 1% variance and the hashed line in the bottom plot shows how many PCs arerequired to explain 80% of the varince in the data. Based on an evaluation of the scree plot we may select 2 or 3 PCs. The cross-validated varince explained (green bars) also suggest that the variance explained does not increase after the first 2 PCs. ##### 2. The sample scores can be used to visualize multivariete similarities in samples given all the varibles for each PC. Lets plot the scores and highlight the sick and healthy groups. ```r #plot results #scores highlighting healthy and sick p.args<-list( pca = res, results = "scores", color = data.cube$sample.meta[,"group",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) ``` ![](report_files/figure-html/unnamed-chunk-5-1.png) #### Based on the scores above the sick and healthy samples look fairly similiar. Lets next look at the variable loadings. #### 3. Variable loadings show the contribution of each varible to the calculated scores. ```r #loadings p.args<-list( pca = res, results = "loadings", color = data.cube$variable.meta[,"type",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) ``` ![](report_files/figure-html/unnamed-chunk-6-1.png) #### Evaluation of the loadings suggest that variance variables X838 abd X454 explain ~90% of the varince in the data. Because we did not scale the data before conducting PCA, variables with the largest magnitude will contribute most to varince explained. #### Next lets recalculate the PCA and mean center and scale all the variables by their standard deviation (autoscale). #### Variance explained ```r args<-list( pca.data = data.cube$data, pca.algorithm = "svd", pca.components = 8, pca.center = TRUE, pca.scaling = "uv", pca.cv = "q2" ) #calculate and view scree plot res2<-devium.pca.calculate(args,return="list",plot=TRUE) ``` ![](report_files/figure-html/unnamed-chunk-7-1.png) #### Variable loadings ```r #loadings p.args<-list( pca = res2, results = "loadings", color = data.cube$variable.meta[,"type",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) ``` ![](report_files/figure-html/unnamed-chunk-8-1.png) #### Sample scores ```r #plot results #scores highlighting healthy and sick p.args<-list( pca = res2, results = "scores", color = data.cube$sample.meta[,"group",drop=FALSE], font.size = 3 ) do.call("plot.PCA",p.args) ``` ![](report_files/figure-html/unnamed-chunk-9-1.png) #### There are some noticible differences in PCA after we scaled our data. 1. Variable magnitude no longer drives the majority of the variance. 2. We can see more resolution in variable loadings for the first 2 PCs. 3. There is an unexplained group structure in the score. #### Next we can try mapping other meta data to score to see if we can explain the cluster pattern. Lets show the analytical batches in the samples scores. ```r p.args<-list( pca = res2, results = "scores", color = data.cube$sample.meta[,"batch",drop=FALSE], font.size =3 ) do.call("plot.PCA",p.args) ``` ![](report_files/figure-html/unnamed-chunk-10-1.png) #### We can see in the scores above that the analytical batch nicely explains 35% of the varince in the data. This is a common problem in large data sets which is best handled using various data normalization methods. Here is some more information about implementing data normalizations. ###### [Metabolomics and Beyond: Challenges and Strategies for Next-generation Omic Analyses](https://imdevsoftware.files.wordpress.com/2015/09/clipboard01.png?w=300&h=225) [![Metabolomics and Beyond: Challenges and Strategies for Next-generation Omic Analyses](https://imdevsoftware.files.wordpress.com/2015/09/clipboard01.png?w=300&h=225)](https://www.youtube.com/watch?v=4AhBN5Q1oMs) ##### [Evaluation of data normalization methods](http://www.slideshare.net/dgrapov/case-study-metabolomic-data-normalization-example) **** #### [Part 3](http://www.slideshare.net/dgrapov/data-analysis-workflows-part-2-2015?related=1) ### Statistical Analysis **** ##### Next lets carry out a statistical analysis and summarise the changes between the sick and ghealthy groups. Below we identify significantly altered analytes using a basic t-test with adjustment for multiple hypotheses tested. We probably want to use more sophisticated and non-parametric tests for real applications. ```r #get summaries and t-test stats data.cube<-data.obj$normalized data<-data.cube$data meta<-data.cube$sample.meta[,"group",drop=FALSE] #get summary .summary<-stats.summary(data,comp.obj=meta,formula=colnames(meta),sigfigs=3,log=FALSE,rel=1,do.stats=TRUE) ``` ``` ## Generating data summary... ## Conducting tests... ## Conducting FDR corrections... ``` ```r stats.obj<-cbind(data.cube$variable.meta,.summary) stats.obj %>% arrange(group_p.values) %>% head(.) ``` ``` ## ID description type ## 1 C00077 ornithine metabolite ## 2 C02477 tocopherol alpha metabolite ## 3 C00097 cysteine metabolite ## 4 C00031 glucose metabolite ## 5 C00170 5'-deoxy-5'-methylthioadenosine metabolite ## 6 C00385 xanthine metabolite ## healthy.mean.....std.dev sick.mean.....std.dev mean.sick_mean.healthy ## 1 7300 +/- 2900 3910 +/- 2100 0.54 ## 2 1160 +/- 1400 2600 +/- 1200 2.24 ## 3 5160 +/- 2500 8460 +/- 3400 1.64 ## 4 335000 +/- 2e+05 144000 +/- 140000 0.43 ## 5 216 +/- 83 355 +/- 160 1.64 ## 6 574 +/- 350 1670 +/- 1500 2.91 ## group_p.values group_adjusted.p.values group_q.values ## 1 9.433317e-06 0.001886663 0.000900412 ## 2 1.505034e-04 0.007714558 0.003632462 ## 3 1.507843e-04 0.007714558 0.003633836 ## 4 1.542912e-04 0.007714558 0.003650646 ## 5 2.004055e-04 0.008016221 0.003825749 ## 6 3.909048e-04 0.013030159 0.006072846 ``` ```r #write.csv(stats.obj,file="results/statistical_results.csv") ``` #### We can visualize the differences in means for the top most altered metabolite and protein as a box plot. ```r top.met<-stats.obj %>% filter(type =="metabolite") %>% arrange(group_p.values) %>% dplyr::select(ID) %>% dplyr:: slice(.,1) %>% unlist(.) %>% as.character(.) id<-as.character(stats.obj$ID) %in% top.met tmp.data<-data.frame(data[,id,drop=FALSE],meta) #make plot ggplot(tmp.data,aes_string(x="group",y=colnames(tmp.data)[1],fill="group")) + geom_boxplot() + ggtitle(as.character(stats.obj$description[id])) +ylab("") ``` ![](report_files/figure-html/unnamed-chunk-12-1.png) ```r top.met<-stats.obj %>% filter(type =="protein") %>% arrange(group_p.values) %>% dplyr::select(ID) %>% dplyr:: slice(.,1) %>% unlist(.) %>% as.character(.) id<-as.character(stats.obj$ID) %in% top.met tmp.data<-data.frame(data[,id,drop=FALSE],meta) #make plot ggplot(tmp.data,aes_string(x="group",y=colnames(tmp.data)[1],fill="group")) + geom_boxplot() + ggtitle(as.character(stats.obj$description[id])) +ylab("") ``` ![](report_files/figure-html/unnamed-chunk-12-2.png) **** ### Predictive Modeling **** #### Next we can try a generate a non-linear multivarite classification model to identify important variables in our data. Below we will train and validate a random forest classifier. The full data set is split into 2/3 trainning and 1/3 test set while keeping the propotion of sick and healthy samples equivalent. The model is trained using 3-fold cross-validation repeated 3 times and the ```mtry``` parameter is optimized to maximize the are under the reciever operator characteristic curve (AUCROC). ```r #create a classification model using random forests #generate training/test set set.seed(998) data<-data.cube$data inTraining <- createDataPartition(data.cube$sample.meta$group, p = 2/3, list = FALSE) train.data <- data[ inTraining,] test.data <- data[-inTraining,] train.y <- data.cube$sample.meta$group[ inTraining] %>% droplevels() test.y <- data.cube$sample.meta$group[ -inTraining] %>% droplevels() #set model parameters fitControl <- trainControl(## 10-fold CV method = "repeatedcv", number = 3, ## repeated ten times repeats = 3, classProbs = TRUE, summaryFunction = twoClassSummary ) #fit model to the training data set.seed(825) fit<- train(train.y ~ ., data = train.data, method = "rf", trControl = fitControl, metric = "ROC", tuneLength = 3 ) ``` #### Below the optimal model is chosen while varying the ```mtry``` or the number of variables randomly sampled as candidates at each split. ```r fit ``` ``` ## Random Forest ## ## 36 samples ## 199 predictors ## 2 classes: 'healthy', 'sick' ## ## No pre-processing ## Resampling: Cross-Validated (3 fold, repeated 3 times) ## ## Summary of sample sizes: 24, 24, 24, 24, 24, 24, ... ## ## Resampling results across tuning parameters: ## ## mtry ROC Sens Spec ROC SD Sens SD Spec SD ## 2 0.7870370 0.7222222 0.7777778 0.08098544 0.2041241 0.1666667 ## 101 0.8549383 0.7777778 0.7592593 0.10090044 0.1443376 0.2060055 ## 200 0.8750000 0.8333333 0.7222222 0.09107554 0.1178511 0.2204793 ## ## ROC was used to select the optimal model using the largest value. ## The final value used for the model was mtry = 200. ``` #### Next we can evaluate the model performance based on predictions for the test set. We can also look at the ROC curve. ```r #predict the test set pred<-predict(fit,newdata=test.data) prob<-predict(fit,newdata=test.data,type="prob") obs<-test.y table(pred,obs) ``` ``` ## obs ## pred healthy sick ## healthy 7 2 ## sick 2 7 ``` ```r #get performance stats twoClassSummary(data=data.frame(obs,pred,prob),lev=levels(pred)) ``` ``` ## ROC Sens Spec ## 0.9135802 0.7777778 0.7777778 ``` #### We can also look at the ROC curve. ```r x<-roc(obs,prob[,levels(pred)[1]],silent = TRUE) plot(x) ``` ![](report_files/figure-html/unnamed-chunk-16-1.png) ``` ## ## Call: ## roc.default(response = obs, predictor = prob[, levels(pred)[1]], silent = TRUE) ## ## Data: prob[, levels(pred)[1]] in 9 controls (obs healthy) > 9 cases (obs sick). ## Area under the curve: 0.9136 ``` #### Having validated our model next we can look at the most important variables driving the classification. We can look at the differences in performance when each variable is randomly permuted or the VIP. ```r #need to get variable names vip<-varImp(fit)$importance # need to keep rownames vip<-vip[order(vip[,1],decreasing=TRUE),,drop=FALSE][1:10,,drop=FALSE] id<-colnames(train.data) %in% gsub('`','',rownames(vip)) tmp.data<-data.frame(importance=vip[,1],variable=factor(stats.obj$description[id],levels=stats.obj$description[id],ordered=FALSE)) #plot ggplot(tmp.data, aes(x=variable,y=importance)) + geom_bar(stat="identity") + coord_flip() ``` ![](report_files/figure-html/unnamed-chunk-17-1.png) ```r id<-as.character(stats.obj$description) %in% as.character(tmp.data[1,2]) tmp.data<-data.frame(data[,id,drop=FALSE],meta) #make plot ggplot(tmp.data,aes_string(x="group",y=colnames(tmp.data)[1],fill="group")) + geom_boxplot() + ggtitle(as.character(stats.obj$description[id])) +ylab("") ``` ![](report_files/figure-html/unnamed-chunk-17-2.png) **** ### Functional Analysis **** #### Finally we can identify enriched biological pathways based on the integrated changes in genes and proteins. [IMPaLA: Integrated Molecular Pathway Level Analysis](http://impala.molgen.mpg.de/) can be used to calculate enriched pathways in genes or proteins and metabolites.To do this we can querry the significantly alterd proteins and metabolites for enriched pathways (see `results/statistical_results_sig.csv`). We can view the full analysis results in `results/IMPaLA_results.csv`. next lets take an enriched pathway and fisualize the fold changes between sick and healthy in the enriched species. ```r #format data to show fold changes in pathway #get formatted data for pathview library(KEGGREST) library(pathview) data<-stats.obj #metabolite met<-data %>% dplyr::filter(type =="metabolite") %>% dplyr::select(ID,mean.sick_mean.healthy) %>% mutate(FC=log(mean.sick_mean.healthy)) %>% dplyr::select(-mean.sick_mean.healthy) #protein prot<-data %>% dplyr::filter(type =="protein") %>% dplyr::select(ID,mean.sick_mean.healthy) %>% mutate(FC=log(mean.sick_mean.healthy)) %>% dplyr::select(-mean.sick_mean.healthy) #set rownames rownames(met)<-met[,1];met<-met[,-1,drop=FALSE] rownames(prot)<-prot[,1];prot<-prot[,-1,drop=FALSE] #select pathway to view path<-"Glycolysis / Gluconeogenesis" ``` #### Lets take a looka at the Glycolysis / Gluconeogenesis pathway. Our data needs to be formatted as below. You can also take a look at the following more detailed example of [mapping fold changes to biochemical pathways](https://github.com/dgrapov/TeachingDemos/blob/master/Demos/Pathway%20Analysis/KEGG%20Pathway%20Enrichment.md). #### Metabolite data showing KEGG ids and log fold change ```r head(met) ``` ``` ## FC ## C00379 0.41871033 ## C00385 1.06815308 ## C00105 0.07696104 ## C00299 -0.24846136 ## C00366 0.33647224 ## C00086 -0.05129329 ``` #### Protein data showing the Entrez gene name and log fold changes ```r head(prot) ``` ``` ## FC ## SPTAN1 -0.3424903 ## CFH 0.1133287 ## VPS13C 0.3148107 ## XRCC6 1.0715836 ## APOA1 -0.1392621 ## SUPT16H 0.7129498 ``` #### Next we need to get the pathway code for or pathway of interest. ```r data(korg) organism <- "homo sapiens" matches <- unlist(sapply(1:ncol(korg), function(i) { agrep(organism, korg[, i]) })) (kegg.code <- korg[matches, 1, drop = F]) ``` ``` ## kegg.code ## [1,] "hsa" ``` #### Now we can visualize the changes between sick and healthy in the Glycolysis / Gluconeogenesis pathway. ```r setwd(wd) pathways <- keggList("pathway", kegg.code) #get code of our pathway of interest map<-grepl(path,pathways) %>% pathways[.] %>% names(.) %>% gsub("path:","",.) map ``` ``` ## [1] "hsa00010" ``` ```r #create image setwd("report") pv.out <- pathview(gene.data = prot, cpd.data = met, gene.idtype = "SYMBOL", pathway.id = map, species = kegg.code, out.suffix = map, keys.align = "y", kegg.native = T, match.data = T, key.pos = "topright") ``` ![](hsa00010.hsa00010.png) **** #### This concludes this short tutorial. You may also find the following links useful. * [Software tools](https://github.com/dgrapov) * [More examples and demos](https://imdevsoftware.wordpress.com/) © Dmitry Grapov (2015) Creative Commons License ================================================ FILE: Demos/Data Analysis Workflow/results/IMPaLA_results.csv ================================================ pathway_name,pathway_source,num_overlapping_genes,overlapping_genes,num_all_pathway_genes,P_genes,Q_genes,num_overlapping_metabolites,overlapping_metabolites,num_all_pathway_metabolites,P_metabolites,Q_metabolites,P_joint,Q_joint Metabolism,Reactome,3,GLS;IVD;ACSS2,1427 (1485),0.505,1,26,C00077;C00158;C00116;C00366;C00385;C00170;C01835;C00245;C00836;C00180;C00315;C00041;C00047;C00106;C00148;C00791;C00020;C00025;C00197;C00097;C00327;C00130;C00135;C03546;C00989;C00346,808 (1017),3.54E-13,8.03E-10,5.43E-12,5.61E-09 Transmembrane transport of small molecules,Reactome,0,,594 (623),1,1,16,C00077;C00106;C00116;C00148;C00245;C00158;C00791;C00020;C00025;C00097;C01571;C00366;C00315;C00041;C00047;C00135,204 (216),4.50E-13,8.03E-10,1.32E-11,6.85E-09 Metabolism of amino acids and derivatives,Reactome,2,IVD;GLS,145 (152),0.0292,1,14,C00077;C00148;C00245;C00791;C00020;C00197;C00025;C00170;C00097;C00315;C00327;C00041;C00047;C00135,181 (190),2.62E-11,1.87E-08,2.21E-11,7.61E-09 Metabolism of amino acids and derivatives,Wikipathways,0,,0 (0),1,1,14,C00077;C00148;C00245;C00791;C00020;C00197;C00025;C00170;C00097;C00315;C00327;C00041;C00047;C00135,174 (187),1.52E-11,1.36E-08,1.52E-11,1.36E-08 SLC-mediated transmembrane transport,Reactome,0,,261 (265),1,1,14,C00077;C00106;C00148;C00135;C00245;C00158;C00791;C00025;C00097;C01571;C00315;C00041;C00047;C00366,160 (165),4.77E-12,5.68E-09,1.29E-10,3.34E-08 Central carbon metabolism in cancer - Homo sapiens (human),KEGG,1,GLS,66 (67),0.116,1,8,C00158;C00148;C00031;C00197;C00025;C00097;C00041;C00135,37 (37),2.36E-10,9.37E-08,6.92E-10,1.43E-07 ABC transporters - Homo sapiens (human),KEGG,0,,44 (44),1,1,12,C00077;C00116;C00148;C00245;C00379;C00031;C00025;C00047;C00315;C00041;C01835;C00135,123 (123),6.60E-11,3.37E-08,1.61E-09,2.39E-07 Amino acid transport across the plasma membrane,Reactome,0,,31 (31),1,1,8,C00077;C00148;C00245;C00025;C00097;C00041;C00047;C00135,32 (32),6.61E-11,3.37E-08,1.62E-09,2.39E-07 Transport of glucose and other sugars_ bile salts and organic acids_ metal ions and amine compounds,Reactome,0,,98 (101),1,1,10,C00158;C00148;C00366;C00791;C00245;C00097;C00315;C00041;C00047;C00135,78 (83),2.28E-10,9.37E-08,5.30E-09,6.86E-07 Urea cycle and metabolism of arginine_ proline_ glutamate_ aspartate and asparagine,EHMN,1,GLS,105 (107),0.178,1,11,C00077;C00148;C00020;C00025;C00170;C00097;C00315;C00327;C00041;C00047;C00989,125 (125),1.52E-09,4.53E-07,6.24E-09,7.18E-07 tRNA charging,HumanCyc,0,,41 (44),1,1,7,C00148;C00020;C00025;C00097;C00041;C00047;C00135,24 (24),3.55E-10,1.27E-07,8.07E-09,7.78E-07 Class I MHC mediated antigen processing & presentation,Reactome,2,SEC24C;SEC13,208 (224),0.0561,1,7,C00148;C00020;C00025;C00097;C00041;C00047;C00135,35 (35),6.48E-09,1.54E-06,8.27E-09,7.78E-07 Glucose Homeostasis,Wikipathways,0,,0 (0),1,1,6,C00077;C00116;C00031;C00327;C00047;C00135,21 (21),8.50E-09,1.78E-06,8.50E-09,1.78E-06 Amino acid and oligopeptide SLC transporters,Reactome,0,,49 (49),1,1,8,C00077;C00148;C00245;C00025;C00097;C00041;C00047;C00135,45 (45),1.26E-09,4.10E-07,2.71E-08,2.34E-06 Transport of inorganic cations/anions and amino acids/oligopeptides,Reactome,0,,94 (95),1,1,8,C00077;C00148;C00245;C00025;C00097;C00041;C00047;C00135,48 (48),2.17E-09,5.97E-07,4.55E-08,3.63E-06 Amino acid synthesis and interconversion (transamination),Reactome,1,GLS,17 (17),0.0311,1,6,C00077;C00148;C00020;C00197;C00025;C00041,32 (32),1.33E-07,1.83E-05,8.42E-08,6.22E-06 Adaptive Immune System,Reactome,3,SEC24C;EVL;SEC13,558 (608),0.0819,1,7,C00148;C00020;C00025;C00097;C00041;C00047;C00135,48 (48),6.59E-08,1.21E-05,1.08E-07,7.46E-06 Endosomal/Vacuolar pathway,Reactome,0,,12 (12),1,1,6,C00148;C00025;C00097;C00041;C00047;C00135,20 (20),6.10E-09,1.54E-06,1.22E-07,7.86E-06 Proton/oligonucleotide cotransporters,Reactome,0,,4 (4),1,1,6,C00148;C00025;C00097;C00041;C00047;C00135,21 (21),8.50E-09,1.78E-06,1.66E-07,1.01E-05 Transport of inorganic cations-anions and amino acids-oligopeptides,Wikipathways,0,,0 (0),1,1,6,C00148;C00025;C00097;C00041;C00047;C00135,31 (32),1.09E-07,1.56E-05,1.09E-07,1.56E-05 γ-glutamyl cycle,HumanCyc,0,,11 (12),1,1,7,C00077;C00148;C00025;C00097;C00041;C00047;C00135,43 (47),2.97E-08,5.89E-06,5.44E-07,3.13E-05 Antigen processing-Cross presentation,Reactome,0,,32 (32),1,1,6,C00148;C00025;C00097;C00041;C00047;C00135,29 (29),7.11E-08,1.21E-05,1.24E-06,6.42E-05 leukotriene biosynthesis,HumanCyc,0,,6 (7),1,1,6,C00148;C00025;C00097;C00041;C00047;C00135,29 (30),7.11E-08,1.21E-05,1.24E-06,6.42E-05 tRNA Aminoacylation,Wikipathways,0,,0 (0),1,1,7,C00148;C00020;C00025;C00097;C00041;C00047;C00135,65 (65),5.67E-07,6.53E-05,5.67E-07,6.53E-05 Glutathione synthesis and recycling,Reactome,0,,14 (17),1,1,6,C00148;C00025;C00097;C00041;C00047;C00135,30 (31),8.83E-08,1.37E-05,1.52E-06,7.16E-05 Urea cycle and metabolism of amino groups,Wikipathways,0,,20 (20),1,1,6,C00077;C00148;C00791;C00025;C00315;C00327,30 (30),8.83E-08,1.37E-05,1.52E-06,7.16E-05 Na+/Cl- dependent neurotransmitter transporters,Reactome,0,,19 (19),1,1,6,C00148;C00245;C00097;C00041;C00047;C00135,31 (31),1.09E-07,1.56E-05,1.85E-06,8.35E-05 Phase II conjugation,Wikipathways,0,,0 (0),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,102 (115),9.38E-07,9.57E-05,9.38E-07,9.57E-05 Nucleotide Metabolism,Wikipathways,0,,19 (19),1,1,5,C00077;C00385;C00130;C00366;C00315,17 (18),1.44E-07,1.90E-05,2.41E-06,0.000104 Amine compound SLC transporters,Reactome,0,,30 (31),1,1,6,C00148;C00245;C00097;C00041;C00047;C00135,35 (35),2.35E-07,2.89E-05,3.82E-06,0.000152 S-methyl-5-thio-α-D-ribose 1-phosphate degradation,HumanCyc,0,,4 (4),1,1,6,C00148;C00025;C00097;C00041;C00047;C00135,35 (35),2.35E-07,2.89E-05,3.82E-06,0.000152 Glutathione conjugation,Reactome,0,,35 (38),1,1,6,C00148;C00025;C00097;C00041;C00047;C00135,38 (41),3.92E-07,4.67E-05,6.18E-06,0.000237 Immune System,Reactome,4,SEC24C;EVL;XRCC6;SEC13,994 (1069),0.106,1,7,C00148;C00020;C00025;C00097;C00041;C00047;C00135,87 (103),4.20E-06,0.000273,6.98E-06,0.000258 Glycine_ serine_ alanine and threonine metabolism,EHMN,1,GLOD4,80 (80),0.139,1,7,C00077;C00020;C00197;C00025;C00097;C00041;C00047,88 (88),4.54E-06,0.000289,9.61E-06,0.000311 Cytosolic tRNA aminoacylation,Reactome,0,,26 (29),1,1,7,C00148;C00020;C00025;C00097;C00041;C00047;C00135,66 (66),6.30E-07,6.62E-05,9.63E-06,0.000311 Mitochondrial tRNA aminoacylation,Reactome,0,,23 (23),1,1,7,C00148;C00020;C00025;C00097;C00041;C00047;C00135,66 (66),6.30E-07,6.62E-05,9.63E-06,0.000311 tRNA Aminoacylation,Reactome,0,,45 (48),1,1,7,C00148;C00020;C00025;C00097;C00041;C00047;C00135,66 (66),6.30E-07,6.62E-05,9.63E-06,0.000311 Metabolism of nucleotides,Wikipathways,0,,0 (0),1,1,7,C00106;C00366;C00385;C00025;C00020;C00130;C00041,97 (111),8.73E-06,0.000528,8.73E-06,0.000528 Chloramphenicol Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Roxithromycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Josamycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Methacycline Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Rolitetracycline Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Streptomycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Spectinomycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Kanamycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Gentamicin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Netilmicin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Neomycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Tobramycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Paromomycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Minocycline Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Tetracycline Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Lincomycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Clindamycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Azithromycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Arbekacin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Tigecycline Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Erythromycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Amikacin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Clomocycline Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Doxycycline Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Demeclocycline Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Oxytetracycline Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Lymecycline Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Telithromycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Clarithromycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Troleandomycin Action Pathway,SMPDB,0,,0 (0),1,1,4,C00020;C00148;C00041;C00135,20 (20),1.55E-05,0.000609,1.55E-05,0.000609 Protein digestion and absorption - Homo sapiens (human),KEGG,0,,87 (89),1,1,6,C00148;C00025;C00097;C00041;C00047;C00135,47 (47),1.45E-06,0.000144,2.09E-05,0.000657 Argininemia,SMPDB,0,,13 (13),1,1,5,C00077;C00327;C00041;C00020;C00025,27 (27),1.77E-06,0.000151,2.52E-05,0.000669 Citrullinemia Type I,SMPDB,0,,13 (13),1,1,5,C00077;C00327;C00041;C00020;C00025,27 (27),1.77E-06,0.000151,2.52E-05,0.000669 Carbamoyl Phosphate Synthetase Deficiency,SMPDB,0,,13 (13),1,1,5,C00077;C00327;C00041;C00020;C00025,27 (27),1.77E-06,0.000151,2.52E-05,0.000669 Argininosuccinic Aciduria,SMPDB,0,,13 (13),1,1,5,C00077;C00327;C00041;C00020;C00025,27 (27),1.77E-06,0.000151,2.52E-05,0.000669 Urea Cycle,SMPDB,0,,13 (13),1,1,5,C00077;C00327;C00041;C00020;C00025,27 (27),1.77E-06,0.000151,2.52E-05,0.000669 Ornithine Transcarbamylase Deficiency (OTC Deficiency),SMPDB,0,,13 (13),1,1,5,C00077;C00327;C00041;C00020;C00025,27 (27),1.77E-06,0.000151,2.52E-05,0.000669 Aminoacyl-tRNA biosynthesis - Homo sapiens (human),KEGG,0,,66 (66),1,1,6,C00148;C00025;C00097;C00041;C00047;C00135,52 (52),2.67E-06,0.000222,3.70E-05,0.000956 Gene Expression,Reactome,3,EEF1D;HNRNPF;RPL9,1182 (1251),0.379,1,7,C00148;C00020;C00025;C00097;C00041;C00047;C00135,97 (104),8.73E-06,0.000528,4.50E-05,0.000991 Phase II conjugation,Reactome,0,,94 (102),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,122 (145),3.69E-06,0.000244,4.98E-05,0.000991 Defective AHCY causes Hypermethioninemia with S-adenosylhomocysteine hydrolase deficiency (HMAHCHD),Reactome,0,,94 (102),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,122 (145),3.69E-06,0.000244,4.98E-05,0.000991 Defective GCLC causes Hemolytic anemia due to gamma-glutamylcysteine synthetase deficiency (HAGGSD),Reactome,0,,94 (102),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,122 (145),3.69E-06,0.000244,4.98E-05,0.000991 Defective UGT1A1 causes hyperbilirubinemia,Reactome,0,,94 (102),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,122 (145),3.69E-06,0.000244,4.98E-05,0.000991 Defective GSS causes Glutathione synthetase deficiency (GSS deficiency),Reactome,0,,94 (102),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,122 (145),3.69E-06,0.000244,4.98E-05,0.000991 Defective GGT1 causes Glutathionuria (GLUTH),Reactome,0,,94 (102),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,122 (145),3.69E-06,0.000244,4.98E-05,0.000991 Defective UGT1A4 causes hyperbilirubinemia,Reactome,0,,94 (102),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,122 (145),3.69E-06,0.000244,4.98E-05,0.000991 Defective MAT1A causes Methionine adenosyltransferase deficiency (MATD),Reactome,0,,94 (102),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,122 (145),3.69E-06,0.000244,4.98E-05,0.000991 Defective SLC35D1 causes Schneckenbecken dysplasia (SCHBCKD),Reactome,0,,94 (102),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,122 (145),3.69E-06,0.000244,4.98E-05,0.000991 Defective TPMT causes Thiopurine S-methyltransferase deficiency (TPMT deficiency),Reactome,0,,94 (102),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,122 (145),3.69E-06,0.000244,4.98E-05,0.000991 Defective OPLAH causes 5-oxoprolinase deficiency (OPLAHD),Reactome,0,,94 (102),1,1,8,C00148;C00020;C00025;C00097;C00180;C00041;C00047;C00135,122 (145),3.69E-06,0.000244,4.98E-05,0.000991 FoxO signaling pathway - Homo sapiens (human),KEGG,0,,127 (134),1,1,3,C00020;C00031;C00025,5 (5),4.73E-06,0.000296,6.27E-05,0.00122 Glycolysis / Gluconeogenesis - Homo sapiens (human),KEGG,3,ADH5;PGK2;ACSS2,66 (67),0.000232,0.303,2,C00031;C00197,31 (31),0.0253,0.229,7.68E-05,0.00147 Metabolism of proteins,Reactome,5,EEF1D;EIF5A2;SEC24C;SEC13;RPL9,662 (693),0.00618,1,6,C00020;C00025;C00170;C01571;C06423;C00315,148 (163),0.000998,0.0166,8.01E-05,0.00151 Arginine and proline metabolism - Homo sapiens (human),KEGG,1,GLS,60 (61),0.106,1,6,C00077;C00148;C00791;C00025;C00315;C00327,91 (91),7.00E-05,0.00227,9.49E-05,0.00175 Disease,Reactome,4,SEC24C;XRCC6;SEC13;ACSS2,1747 (1852),0.412,1,14,C00077;C00158;C00148;C00020;C00197;C01835;C00025;C00097;C00180;C00315;C00327;C00041;C00047;C00135,533 (699),2.61E-05,0.00099,0.000133,0.00242 Biological oxidations,Reactome,1,ACSS2,174 (189),0.278,1,9,C00148;C00020;C00025;C00097;C00180;C00315;C00041;C00047;C00135,220 (278),3.96E-05,0.00147,0.000137,0.00244 One carbon donor,Wikipathways,0,,1 (1),1,1,4,C00077;C00245;C00097;C00315,19 (23),1.25E-05,0.000609,0.000153,0.00265 purine nucleotides degradation,HumanCyc,0,,12 (12),1,1,4,C00385;C00020;C00130;C00366,19 (19),1.25E-05,0.000609,0.000153,0.00265 Amino acid conjugation of benzoic acid,Wikipathways,1,ACSS2,4 (4),0.00741,1,2,C00020;C00180,9 (9),0.00219,0.0312,0.000195,0.00331 Urea cycle,Reactome,0,,9 (10),1,1,4,C00077;C00327;C00020;C00025,21 (21),1.91E-05,0.00074,0.000226,0.00378 Taurine and hypotaurine metabolism - Homo sapiens (human),KEGG,0,,10 (10),1,1,4,C00041;C00245;C00097;C00025,22 (22),2.32E-05,0.000889,0.00027,0.00444 Metabolism of carbohydrates,Wikipathways,0,,0 (0),1,1,5,C00158;C00020;C00197;C01835;C00025,66 (69),0.000157,0.00459,0.000157,0.00459 Metabolism of nucleotides,Reactome,0,,81 (83),1,1,7,C00106;C00366;C00385;C00025;C00020;C00130;C00041,122 (135),3.95E-05,0.00147,0.00044,0.00712 Gamma carboxylation_ hypusine formation and arylsulfatase activation,Reactome,1,EIF5A2,35 (35),0.0631,1,3,C00020;C00170;C00315,23 (23),0.000758,0.0135,0.000524,0.00725 Hyperornithinemia with gyrate atrophy (HOGA),SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 Creatine deficiency_ guanidinoacetate methyltransferase deficiency,SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 L-arginine:glycine amidinotransferase deficiency,SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 Hyperornithinemia-hyperammonemia-homocitrullinuria [HHH-syndrome],SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 Guanidinoacetate Methyltransferase Deficiency (GAMT Deficiency),SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 Prolinemia Type II,SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 Prolidase Deficiency (PD),SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 Hyperprolinemia Type I,SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 Hyperprolinemia Type II,SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 Ornithine Aminotransferase Deficiency (OAT Deficiency),SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 Arginine: Glycine Amidinotransferase Deficiency (AGAT Deficiency),SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 Arginine and Proline Metabolism,SMPDB,0,,20 (20),1,1,5,C00327;C00077;C00148;C00020;C00025,52 (52),4.94E-05,0.00163,0.000539,0.00725 molybdenum cofactor biosynthesis,HumanCyc,0,,4 (4),1,1,3,C00020;C00041;C00097,10 (13),5.52E-05,0.00181,0.000596,0.00791 Metabolic disorders of biological oxidation enzymes,Reactome,1,ACSS2,584 (641),0.672,1,10,C00077;C00148;C00020;C00025;C00097;C00180;C00315;C00041;C00047;C00135,305 (382),9.00E-05,0.00285,0.000649,0.0085 Metabolism of water-soluble vitamins and cofactors,Wikipathways,0,,0 (0),1,1,5,C00020;C00041;C00047;C00097;C00025,83 (96),0.000462,0.00988,0.000462,0.00988 urate biosynthesis/inosine 5_-phosphate degradation,HumanCyc,0,,6 (6),1,1,3,C00385;C00130;C00366,11 (11),7.55E-05,0.00241,0.000792,0.0101 Glucose-Alanine Cycle,SMPDB,0,,8 (8),1,1,3,C00031;C00041;C00025,11 (11),7.55E-05,0.00241,0.000792,0.0101 Alanine_ aspartate and glutamate metabolism - Homo sapiens (human),KEGG,1,GLS,35 (35),0.0631,1,3,C00158;C00041;C00025,28 (28),0.00136,0.0214,0.000892,0.0113 Synthesis_ Secretion_ and Deacylation of Ghrelin,Wikipathways,0,,0 (0),1,1,2,C01571;C06423,5 (5),0.000621,0.0121,0.000621,0.0121 triacylglycerol degradation,HumanCyc,0,,15 (15),1,1,4,C01571;C00116;C06423;C01601,32 (56),0.000108,0.00334,0.00109,0.0134 Purine catabolism,Reactome,0,,11 (11),1,1,4,C00385;C00020;C00130;C00366,32 (32),0.000108,0.00334,0.00109,0.0134 COPII (Coat Protein 2) Mediated Vesicle Transport,Reactome,2,SEC24C;SEC13,9 (10),0.000117,0.228,0,,3 (3),1,1,0.00118,0.0142 ER to Golgi Transport,Reactome,2,SEC24C;SEC13,9 (10),0.000117,0.228,0,,3 (3),1,1,0.00118,0.0142 Gluconeogenesis,Reactome,0,,32 (33),1,1,4,C00158;C00020;C00197;C00025,33 (33),0.000122,0.00375,0.00122,0.0145 adenosine nucleotides degradation,HumanCyc,0,,9 (9),1,1,3,C00385;C00020;C00366,13 (13),0.000129,0.00395,0.00129,0.0151 SREBP signalling,Wikipathways,2,SEC24C;SEC13,60 (60),0.00543,1,1,C00031,3 (5),0.024,0.219,0.0013,0.0151 Hypoacetylaspartia,SMPDB,0,,14 (14),1,1,4,C00020;C00327;C00130;C00025,34 (34),0.000137,0.00408,0.00136,0.0153 Canavan Disease,SMPDB,0,,14 (14),1,1,4,C00020;C00327;C00130;C00025,34 (34),0.000137,0.00408,0.00136,0.0153 Aspartate Metabolism,SMPDB,0,,14 (14),1,1,4,C00020;C00327;C00130;C00025,34 (34),0.000137,0.00408,0.00136,0.0153 [2Fe-2S] iron-sulfur cluster biosynthesis,HumanCyc,0,,0 (0),1,1,2,C00041;C00097,6 (6),0.000927,0.0155,0.000927,0.0155 Purine metabolism,Reactome,0,,34 (36),1,1,5,C00385;C00020;C00130;C00366;C00025,66 (67),0.000157,0.00459,0.00153,0.017 Trans-sulfuration pathway,Wikipathways,0,,10 (10),1,1,3,C00245;C00097;C00025,14 (14),0.000164,0.00476,0.00159,0.0175 GPCR downstream signaling,Wikipathways,0,,0 (0),1,1,5,C00077;C00020;C00116;C00047;C00025,101 (121),0.00114,0.0183,0.00114,0.0183 Post-translational protein modification,Reactome,3,EIF5A2;SEC24C;SEC13,281 (290),0.0144,1,4,C00020;C00025;C00170;C00315,113 (126),0.0122,0.129,0.00169,0.0184 glutamine degradation/glutamate biosynthesis,HumanCyc,1,GLS,3 (3),0.00556,1,1,C00025,4 (4),0.0319,0.26,0.00171,0.0184 Synthesis_ secretion_ and deacylation of Ghrelin,Reactome,0,,8 (9),1,1,2,C01571;C06423,3 (3),0.000188,0.00542,0.0018,0.0192 guanosine ribonucleotides de novo biosynthesis,HumanCyc,0,,13 (13),1,1,3,C00020;C00130;C00025,15 (15),0.000204,0.00577,0.00193,0.0201 Molybdenum cofactor biosynthesis,Reactome,0,,6 (6),1,1,3,C00020;C00041;C00097,15 (20),0.000204,0.00577,0.00193,0.0201 Glutathione metabolism - Homo sapiens (human),KEGG,0,,51 (51),1,1,4,C00077;C00025;C00097;C00315,38 (38),0.000213,0.006,0.00202,0.0201 Xanthine Dehydrogenase Deficiency (Xanthinuria),SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Adenylosuccinate Lyase Deficiency,SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 AICA-Ribosiduria,SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Adenine phosphoribosyltransferase deficiency (APRT),SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Purine Metabolism,SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Molybdenum Cofactor Deficiency,SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Adenosine Deaminase Deficiency,SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Gout or Kelley-Seegmiller Syndrome,SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Lesch-Nyhan Syndrome (LNS),SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Xanthinuria type I,SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Xanthinuria type II,SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Purine Nucleoside Phosphorylase Deficiency,SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Mitochondrial DNA depletion syndrome,SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 Myoadenylate deaminase deficiency,SMPDB,0,,37 (37),1,1,5,C00385;C00020;C00130;C00025;C00366,72 (72),0.000237,0.00601,0.00222,0.0201 sphingosine and sphingosine-1-phosphate metabolism,HumanCyc,0,,9 (9),1,1,4,C01571;C06423;C01601;C00346,40 (59),0.000261,0.00657,0.00242,0.0217 acetate conversion to acetyl-CoA,HumanCyc,1,ACSS2,3 (3),0.00556,1,1,C00020,6 (6),0.0475,0.336,0.00244,0.0217 Methionine Adenosyltransferase Deficiency,SMPDB,0,,19 (20),1,1,4,C00020;C00170;C00097;C00315,41 (42),0.000288,0.00685,0.00263,0.0217 Glycine N-methyltransferase Deficiency,SMPDB,0,,19 (20),1,1,4,C00020;C00170;C00097;C00315,41 (42),0.000288,0.00685,0.00263,0.0217 Hypermethioninemia,SMPDB,0,,19 (20),1,1,4,C00020;C00170;C00097;C00315,41 (42),0.000288,0.00685,0.00263,0.0217 Methylenetetrahydrofolate Reductase Deficiency (MTHFRD),SMPDB,0,,19 (20),1,1,4,C00020;C00170;C00097;C00315,41 (42),0.000288,0.00685,0.00263,0.0217 Homocystinuria-megaloblastic anemia due to defect in cobalamin metabolism_ cblG complementation type,SMPDB,0,,19 (20),1,1,4,C00020;C00170;C00097;C00315,41 (42),0.000288,0.00685,0.00263,0.0217 Cystathionine Beta-Synthase Deficiency,SMPDB,0,,19 (20),1,1,4,C00020;C00170;C00097;C00315,41 (42),0.000288,0.00685,0.00263,0.0217 S-Adenosylhomocysteine (SAH) Hydrolase Deficiency,SMPDB,0,,19 (20),1,1,4,C00020;C00170;C00097;C00315,41 (42),0.000288,0.00685,0.00263,0.0217 Methionine Metabolism,SMPDB,0,,19 (20),1,1,4,C00020;C00170;C00097;C00315,41 (42),0.000288,0.00685,0.00263,0.0217 urea cycle,HumanCyc,0,,6 (7),1,1,3,C00327;C00077;C00020,17 (17),0.000301,0.00685,0.00274,0.0217 Pyruvate Carboxylase Deficiency,SMPDB,0,,5 (5),1,1,3,C00020;C00041;C00025,17 (17),0.000301,0.00685,0.00274,0.0217 Primary Hyperoxaluria Type I,SMPDB,0,,5 (5),1,1,3,C00020;C00041;C00025,17 (17),0.000301,0.00685,0.00274,0.0217 Alanine Metabolism,SMPDB,0,,5 (5),1,1,3,C00020;C00041;C00025,17 (17),0.000301,0.00685,0.00274,0.0217 Spermidine and Spermine Biosynthesis,SMPDB,0,,6 (6),1,1,3,C00077;C00170;C00315,17 (17),0.000301,0.00685,0.00274,0.0217 Lactic Acidemia,SMPDB,0,,5 (5),1,1,3,C00020;C00041;C00025,17 (17),0.000301,0.00685,0.00274,0.0217 guanosine nucleotides de novo biosynthesis,HumanCyc,0,,16 (16),1,1,3,C00020;C00130;C00025,17 (17),0.000301,0.00685,0.00274,0.0217 Gastrin-CREB signalling pathway via PKC and MAPK,Reactome,0,,203 (222),1,1,4,C00077;C00116;C00047;C00025,42 (46),0.000316,0.0071,0.00287,0.0223 G alpha (q) signalling events,Reactome,0,,175 (190),1,1,4,C00077;C00116;C00047;C00025,42 (46),0.000316,0.0071,0.00287,0.0223 Transport of glucose and other sugars_ bile salts and organic acids_ metal ions and amine compounds,Wikipathways,0,,0 (0),1,1,3,C00158;C00148;C00366,29 (29),0.00151,0.0232,0.00151,0.0232 Glucose metabolism,Reactome,0,,67 (70),1,1,4,C00158;C00020;C00197;C00025,43 (43),0.000347,0.00774,0.00311,0.024 Hypusine synthesis from eIF5A-lysine,Reactome,1,EIF5A2,4 (4),0.00741,1,1,C00315,6 (6),0.0475,0.336,0.00315,0.0241 thio-molybdenum cofactor biosynthesis,HumanCyc,0,,1 (1),1,1,2,C00041;C00097,4 (6),0.000375,0.00821,0.00333,0.025 alanine biosynthesis/degradation,HumanCyc,0,,2 (2),1,1,2,C00041;C00025,4 (4),0.000375,0.00821,0.00333,0.025 Proton-coupled neutral amino acid transporters,Reactome,0,,2 (2),1,1,2,C00041;C00148,4 (4),0.000375,0.00821,0.00333,0.025 Methionine and cysteine metabolism,EHMN,0,,79 (80),1,1,5,C00020;C00041;C00245;C00097;C00025,80 (80),0.000389,0.00847,0.00344,0.0256 Vitamin B12 Metabolism,Wikipathways,0,,50 (51),1,1,4,C00031;C00097;C02477;C00791,46 (59),0.000451,0.00976,0.00392,0.029 Mercaptopurine Action Pathway,SMPDB,0,,47 (47),1,1,5,C00020;C00385;C00130;C00025;C00366,83 (83),0.000462,0.00988,0.00401,0.0294 Azathioprine Action Pathway,SMPDB,0,,47 (47),1,1,5,C00020;C00385;C00130;C00025;C00366,84 (84),0.000488,0.0103,0.00421,0.0305 Thioguanine Action Pathway,SMPDB,0,,47 (47),1,1,5,C00020;C00385;C00130;C00025;C00366,84 (84),0.000488,0.0103,0.00421,0.0305 Mitochondrial Iron-Sulfur Cluster Biogenesis,Wikipathways,0,,0 (0),1,1,2,C00041;C00097,9 (12),0.00219,0.0312,0.00219,0.0312 Selenium Micronutrient Network,Wikipathways,0,,76 (83),1,1,5,C00385;C00031;C02477;C00097;C00366,85 (104),0.000516,0.0108,0.00442,0.0313 2-Hydroxyglutric Aciduria (D And L Form),SMPDB,0,,22 (23),1,1,4,C00020;C00041;C00025;C00097,48 (48),0.000531,0.0108,0.00454,0.0313 Homocarnosinosis,SMPDB,0,,22 (23),1,1,4,C00020;C00041;C00025;C00097,48 (48),0.000531,0.0108,0.00454,0.0313 Succinic semialdehyde dehydrogenase deficiency,SMPDB,0,,22 (23),1,1,4,C00020;C00041;C00025;C00097,48 (48),0.000531,0.0108,0.00454,0.0313 4-Hydroxybutyric Aciduria/Succinic Semialdehyde Dehydrogenase Deficiency,SMPDB,0,,22 (23),1,1,4,C00020;C00041;C00025;C00097,48 (48),0.000531,0.0108,0.00454,0.0313 Glutamate Metabolism,SMPDB,0,,22 (23),1,1,4,C00020;C00041;C00025;C00097,48 (48),0.000531,0.0108,0.00454,0.0313 Hyperinsulinism-Hyperammonemia Syndrome,SMPDB,0,,22 (23),1,1,4,C00020;C00041;C00025;C00097,48 (48),0.000531,0.0108,0.00454,0.0313 Vitamin E metabolism,EHMN,0,,43 (43),1,1,3,C00020;C00047;C02477,21 (21),0.000576,0.0116,0.00487,0.0334 NAD Biosynthesis II (from tryptophan),Wikipathways,0,,8 (8),1,1,3,C00020;C00041;C00025,22 (23),0.000663,0.0121,0.00552,0.0336 Defective MMACHC causes methylmalonic aciduria and homocystinuria type cblC,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Metabolism of vitamins and cofactors,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective GIF causes intrinsic factor deficiency,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective AMN causes hereditary megaloblastic anemia 1,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective MMAB causes methylmalonic aciduria type cblB,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective MMAA causes methylmalonic aciduria type cblA,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defects in vitamin and cofactor metabolism,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective MUT causes methylmalonic aciduria mut type,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Metabolism of water-soluble vitamins and cofactors,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective TCN2 causes hereditary megaloblastic anemia,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective CD320 causes methylmalonic aciduria,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective MTRR causes methylmalonic aciduria and homocystinuria type cblE,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defects in cobalamin (B12) metabolism,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective HLCS causes multiple carboxylase deficiency,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective BTD causes biotidinase deficiency,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defects in biotin (Btn) metabolism,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective LMBRD1 causes methylmalonic aciduria and homocystinuria type cblF,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective CUBN causes hereditary megaloblastic anemia 1,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective MTR causes methylmalonic aciduria and homocystinuria type cblG,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 Defective MMADHC causes methylmalonic aciduria and homocystinuria type cblD,Reactome,0,,83 (88),1,1,5,C00020;C00041;C00025;C00097;C00047,90 (101),0.000672,0.0121,0.00558,0.0336 D-Glutamine and D-glutamate metabolism - Homo sapiens (human),KEGG,1,GLS,4 (4),0.00741,1,1,C00025,12 (12),0.0927,0.469,0.00569,0.034 Sulfur amino acid metabolism,Reactome,0,,23 (25),1,1,4,C00097;C00025;C00170;C00245,52 (59),0.000723,0.0129,0.00595,0.0354 Antigen Presentation: Folding_ assembly and peptide loading of class I MHC,Reactome,2,SEC24C;SEC13,23 (24),0.000809,0.79,0,,7 (7),1,1,0.00657,0.0389 Purine metabolism,EHMN,0,,220 (228),1,1,5,C00385;C00020;C00130;C00025;C00366,94 (94),0.00082,0.0145,0.00665,0.0391 Gastrin-CREB signalling pathway via PKC and MAPK,Wikipathways,0,,0 (0),1,1,3,C00077;C00047;C00025,36 (40),0.00285,0.0393,0.00285,0.0393 Glycolysis,Reactome,0,,28 (29),1,1,3,C00158;C00020;C00197,24 (24),0.000862,0.0151,0.00694,0.0399 NAD de novo biosynthesis,HumanCyc,0,,12 (15),1,1,3,C00020;C00041;C00025,24 (24),0.000862,0.0151,0.00694,0.0399 3-Phosphoglycerate dehydrogenase deficiency,SMPDB,0,,26 (27),1,1,4,C00020;C00041;C00025;C00097,55 (55),0.000895,0.0151,0.00718,0.0399 Non Ketotic Hyperglycinemia,SMPDB,0,,26 (27),1,1,4,C00020;C00041;C00025;C00097,55 (55),0.000895,0.0151,0.00718,0.0399 Glycine and Serine Metabolism,SMPDB,0,,26 (27),1,1,4,C00020;C00041;C00025;C00097,55 (55),0.000895,0.0151,0.00718,0.0399 Dimethylglycinuria,SMPDB,0,,26 (27),1,1,4,C00020;C00041;C00025;C00097,55 (55),0.000895,0.0151,0.00718,0.0399 Hyperglycinemia_ non-ketotic,SMPDB,0,,26 (27),1,1,4,C00020;C00041;C00025;C00097,55 (55),0.000895,0.0151,0.00718,0.0399 Dimethylglycine Dehydrogenase Deficiency,SMPDB,0,,26 (27),1,1,4,C00020;C00041;C00025;C00097,55 (55),0.000895,0.0151,0.00718,0.0399 Dihydropyrimidine Dehydrogenase Deficiency (DHPD),SMPDB,0,,26 (27),1,1,4,C00020;C00041;C00025;C00097,55 (55),0.000895,0.0151,0.00718,0.0399 Sarcosinemia,SMPDB,0,,26 (27),1,1,4,C00020;C00041;C00025;C00097,55 (55),0.000895,0.0151,0.00718,0.0399 Glycolysis and Gluconeogenesis,EHMN,1,PGK2,68 (70),0.119,1,3,C00158;C00020;C00197,52 (52),0.00807,0.0921,0.00764,0.0423 Folate-Alcohol and Cancer Pathway,Wikipathways,1,ADH5,8 (8),0.0148,1,1,C00097,9 (10),0.0703,0.401,0.00817,0.045 Metabolism of carbohydrates,Reactome,0,,256 (261),1,1,5,C00158;C00020;C00197;C01835;C00025,100 (120),0.00109,0.0177,0.0085,0.0454 Myoclonic epilepsy of Lafora,Reactome,0,,256 (261),1,1,5,C00158;C00020;C00197;C01835;C00025,100 (120),0.00109,0.0177,0.0085,0.0454 Glycogen storage diseases,Reactome,0,,256 (261),1,1,5,C00158;C00020;C00197;C01835;C00025,100 (120),0.00109,0.0177,0.0085,0.0454 superpathway of tryptophan utilization,HumanCyc,0,,43 (47),1,1,4,C00020;C00041;C00025;C00097,58 (70),0.00109,0.0177,0.00856,0.0454 Beta-mercaptolactate-cysteine disulfiduria,SMPDB,0,,9 (9),1,1,3,C00020;C00025;C00097,26 (26),0.0011,0.0177,0.00856,0.0454 Cysteine Metabolism,SMPDB,0,,9 (9),1,1,3,C00020;C00025;C00097,26 (26),0.0011,0.0177,0.00856,0.0454 Cystinosis_ ocular nonnephropathic,SMPDB,0,,9 (9),1,1,3,C00020;C00025;C00097,26 (26),0.0011,0.0177,0.00856,0.0454 ethanol degradation IV,HumanCyc,1,ACSS2,6 (6),0.0111,1,1,C00020,13 (13),0.1,0.491,0.00866,0.0457 fatty acid β-oxidation,HumanCyc,0,,19 (20),1,1,4,C00020;C01571;C06423;C01601,60 (85),0.00124,0.0199,0.00957,0.0503 Synthesis of diphthamide-EEF2,Reactome,0,,8 (8),1,1,2,C00020;C00170,7 (7),0.00129,0.0204,0.00988,0.0511 spermidine biosynthesis,HumanCyc,0,,2 (2),1,1,2,C00170;C00315,7 (7),0.00129,0.0204,0.00988,0.0511 spermine biosynthesis,HumanCyc,0,,2 (2),1,1,2,C00170;C00315,7 (7),0.00129,0.0204,0.00988,0.0511 fatty acid β-oxidation (peroxisome),HumanCyc,0,,19 (20),1,1,4,C00020;C01571;C06423;C01601,61 (85),0.00132,0.0208,0.0101,0.052 ethanol degradation II,HumanCyc,1,ACSS2,8 (8),0.0148,1,1,C00020,12 (12),0.0927,0.469,0.0104,0.0532 GPCR downstream signaling,Reactome,0,,930 (986),1,1,5,C00077;C00020;C00116;C00025;C00047,107 (126),0.00147,0.023,0.0111,0.0565 oxidative ethanol degradation III,HumanCyc,1,ACSS2,7 (7),0.0129,1,1,C00020,15 (15),0.115,0.54,0.0111,0.0565 acyl-CoA hydrolysis,HumanCyc,0,,3 (3),1,1,3,C01571;C06423;C01601,29 (39),0.00151,0.0232,0.0113,0.0567 Mineral absorption - Homo sapiens (human),KEGG,0,,50 (51),1,1,3,C00148;C00031;C00041,29 (29),0.00151,0.0232,0.0113,0.0567 Ammonia Recycling,SMPDB,0,,12 (12),1,1,3,C00020;C00025;C00135,29 (29),0.00151,0.0232,0.0113,0.0567 superpathway of purine nucleotide salvage,HumanCyc,0,,59 (61),1,1,3,C00020;C00130;C00025,30 (30),0.00167,0.0254,0.0124,0.0613 Metabolism of polyamines,Reactome,0,,14 (14),1,1,3,C00077;C00170;C00315,30 (34),0.00167,0.0254,0.0124,0.0613 Conjugation of benzoate with glycine,Reactome,0,,5 (6),1,1,2,C00020;C00180,8 (8),0.00171,0.0256,0.0126,0.0614 lipoate biosynthesis and incorporation,HumanCyc,0,,3 (3),1,1,2,C00020;C06423,8 (8),0.00171,0.0256,0.0126,0.0614 L-cysteine degradation II,HumanCyc,0,,2 (2),1,1,2,C00025;C00097,8 (8),0.00171,0.0256,0.0126,0.0614 Biotin transport and metabolism,Reactome,0,,11 (11),1,1,2,C00020;C00047,8 (8),0.00171,0.0256,0.0126,0.0614 Glutamate Neurotransmitter Release Cycle,Reactome,1,GLS,20 (21),0.0365,1,1,C00025,6 (6),0.0475,0.336,0.0127,0.0617 Ethanol Degradation,SMPDB,1,ACSS2,7 (7),0.0129,1,1,C00020,18 (18),0.136,0.616,0.0129,0.0621 Membrane Trafficking,Reactome,2,SEC24C;SEC13,153 (162),0.0322,1,1,C00020,7 (7),0.0551,0.361,0.013,0.0624 beta-Alanine metabolism - Homo sapiens (human),KEGG,0,,30 (30),1,1,3,C00106;C00135;C00315,31 (31),0.00184,0.0274,0.0134,0.0638 Ethanol oxidation,Reactome,1,ACSS2,10 (10),0.0184,1,1,C00020,13 (13),0.1,0.491,0.0134,0.0638 Transport to the Golgi and subsequent modification,Reactome,2,SEC24C;SEC13,36 (37),0.00199,1,0,,23 (24),1,1,0.0143,0.0678 sphingomyelin metabolism/ceramide salvage,HumanCyc,0,,8 (8),1,1,3,C01571;C06423;C01601,32 (55),0.00202,0.03,0.0146,0.0685 glutathione biosynthesis,HumanCyc,0,,3 (4),1,1,2,C00025;C00097,9 (9),0.00219,0.0312,0.0156,0.0708 L-glutamine tRNA biosynthesis,HumanCyc,0,,2 (2),1,1,2,C00020;C00025,9 (9),0.00219,0.0312,0.0156,0.0708 diphthamide biosynthesis,HumanCyc,0,,2 (2),1,1,2,C00020;C00170,9 (9),0.00219,0.0312,0.0156,0.0708 proline degradation,HumanCyc,0,,2 (2),1,1,2,C00148;C00025,9 (10),0.00219,0.0312,0.0156,0.0708 Carnosinuria_ carnosinemia,SMPDB,0,,9 (9),1,1,3,C00106;C00025;C00135,33 (34),0.00221,0.0312,0.0157,0.0708 Ureidopropionase deficiency,SMPDB,0,,9 (9),1,1,3,C00106;C00025;C00135,33 (34),0.00221,0.0312,0.0157,0.0708 GABA-Transaminase Deficiency,SMPDB,0,,9 (9),1,1,3,C00106;C00025;C00135,33 (34),0.00221,0.0312,0.0157,0.0708 Organic cation/anion/zwitterion transport,Reactome,0,,13 (13),1,1,3,C00315;C00366;C00791,33 (36),0.00221,0.0312,0.0157,0.0708 Beta-Alanine Metabolism,SMPDB,0,,9 (9),1,1,3,C00106;C00025;C00135,33 (34),0.00221,0.0312,0.0157,0.0708 phospholipases,HumanCyc,0,,41 (42),1,1,3,C01571;C06423;C01601,33 (55),0.00221,0.0312,0.0157,0.0708 superpathway of conversion of glucose to acetyl CoA and entry into the TCA cycle,HumanCyc,1,PGK2,48 (52),0.0855,1,2,C00158;C00197,34 (36),0.0301,0.26,0.0179,0.0804 Pentose phosphate pathway - Homo sapiens (human),KEGG,0,,28 (28),1,1,3,C00031;C00197;C00257,35 (35),0.00262,0.0369,0.0182,0.0813 Sulfur relay system - Homo sapiens (human),KEGG,0,,10 (10),1,1,2,C00041;C00097,10 (10),0.00273,0.0377,0.0188,0.0826 Proline catabolism,Reactome,0,,2 (2),1,1,2,C00148;C00025,10 (10),0.00273,0.0377,0.0188,0.0826 L-cysteine degradation I,HumanCyc,0,,2 (2),1,1,2,C00025;C00097,10 (10),0.00273,0.0377,0.0188,0.0826 actions of nitric oxide in the heart,BioCarta,0,,41 (43),1,1,2,C00327;C00020,10 (10),0.00273,0.0377,0.0188,0.0826 Cori Cycle,Wikipathways,1,PGK2,15 (15),0.0275,1,1,C00031,13 (15),0.1,0.491,0.019,0.0829 Triglyceride Biosynthesis,Reactome,0,,43 (45),1,1,3,C00158;C00020;C00116,37 (40),0.00308,0.0422,0.0209,0.0905 Purine ribonucleoside monophosphate biosynthesis,Reactome,0,,11 (12),1,1,3,C00020;C00130;C00025,37 (37),0.00308,0.0422,0.0209,0.0905 Abacavir transport and metabolism,Wikipathways,0,,0 (0),1,1,2,C00020;C00130,17 (24),0.00795,0.0916,0.00795,0.0916 leucine degradation,HumanCyc,1,IVD,12 (12),0.0221,1,1,C00025,20 (20),0.15,0.653,0.0222,0.092 DNA Repair,Reactome,1,XRCC6,117 (120),0.196,1,2,C00020;C00106,25 (32),0.0169,0.169,0.0222,0.092 serine biosynthesis (phosphorylated route),HumanCyc,0,,4 (4),1,1,2,C00197;C00025,11 (11),0.00332,0.0438,0.0222,0.092 asparagine biosynthesis,HumanCyc,0,,3 (3),1,1,2,C00020;C00025,11 (11),0.00332,0.0438,0.0222,0.092 taurine biosynthesis,HumanCyc,0,,3 (3),1,1,2,C00245;C00097,11 (11),0.00332,0.0438,0.0222,0.092 Phosphate bond hydrolysis by NUDT proteins,Reactome,0,,6 (6),1,1,2,C00020;C00130,11 (22),0.00332,0.0438,0.0222,0.092 Sphingolipid Metabolism,SMPDB,0,,22 (22),1,1,3,C00031;C00836;C00346,38 (39),0.00333,0.0438,0.0223,0.092 Gaucher Disease,SMPDB,0,,22 (22),1,1,3,C00031;C00836;C00346,38 (39),0.00333,0.0438,0.0223,0.092 Globoid Cell Leukodystrophy,SMPDB,0,,22 (22),1,1,3,C00031;C00836;C00346,38 (39),0.00333,0.0438,0.0223,0.092 Metachromatic Leukodystrophy (MLD),SMPDB,0,,22 (22),1,1,3,C00031;C00836;C00346,38 (39),0.00333,0.0438,0.0223,0.092 Fabry disease,SMPDB,0,,22 (22),1,1,3,C00031;C00836;C00346,38 (39),0.00333,0.0438,0.0223,0.092 Krabbe disease,SMPDB,0,,22 (22),1,1,3,C00031;C00836;C00346,38 (39),0.00333,0.0438,0.0223,0.092 Transport of Glycerol from Adipocytes to the Liver by Aquaporins,Wikipathways,0,,0 (0),1,1,1,C00116,1 (1),0.00806,0.0921,0.00806,0.0921 Histidine Metabolism,SMPDB,0,,15 (15),1,1,3,C00020;C00025;C00135,39 (40),0.00358,0.0467,0.0238,0.0969 Histidinemia,SMPDB,0,,15 (15),1,1,3,C00020;C00025;C00135,39 (40),0.00358,0.0467,0.0238,0.0969 the visual cycle I (vertebrates),HumanCyc,0,,17 (18),1,1,3,C01571;C06423;C01601,39 (58),0.00358,0.0467,0.0238,0.0969 Glycerophospholipid Biosynthetic Pathway,Wikipathways,0,,0 (0),1,1,2,C00116;C00346,18 (25),0.0089,0.0984,0.0089,0.0984 formaldehyde oxidation,HumanCyc,1,ADH5,2 (2),0.00371,1,0,,11 (11),1,1,0.0245,0.0993 purine nucleotides de novo biosynthesis,HumanCyc,0,,59 (62),1,1,3,C00020;C00130;C00025,40 (40),0.00385,0.05,0.0253,0.102 Alpha9 beta1 integrin signaling events,PID,0,,24 (26),1,1,2,C00327;C00315,12 (13),0.00396,0.0504,0.0259,0.102 guanosine nucleotides degradation,HumanCyc,0,,4 (4),1,1,2,C00385;C00366,12 (12),0.00396,0.0504,0.0259,0.102 Taurine and Hypotaurine Metabolism,SMPDB,0,,5 (5),1,1,2,C00245;C00097,12 (12),0.00396,0.0504,0.0259,0.102 chrebp regulation by carbohydrates and camp,BioCarta,0,,40 (42),1,1,2,C00020;C00031,12 (12),0.00396,0.0504,0.0259,0.102 adenosine ribonucleotides de novo biosynthesis,HumanCyc,0,,38 (40),1,1,2,C00020;C00130,12 (12),0.00396,0.0504,0.0259,0.102 Signaling by GPCR,Reactome,0,,1046 (1108),1,1,5,C00077;C00020;C00116;C00025;C00047,134 (153),0.00396,0.0504,0.0259,0.102 Valproic Acid Metabolism Pathway,SMPDB,1,IVD,11 (11),0.0202,1,1,C00020,29 (29),0.21,0.847,0.0275,0.108 Signal Transduction,Reactome,0,,1966 (2077),1,1,6,C00077;C00116;C00020;C00025;C00327;C00047,198 (231),0.0044,0.0558,0.0283,0.111 Class C/3 (Metabotropic glutamate/pheromone receptors),Reactome,0,,36 (41),1,1,3,C00077;C00047;C00025,42 (54),0.00443,0.0559,0.0284,0.111 Stimuli-sensing channels,Reactome,0,,94 (102),1,1,2,C00020;C00366,13 (14),0.00465,0.0577,0.0296,0.114 Non-homologous end-joining - Homo sapiens (human),KEGG,1,XRCC6,13 (13),0.0239,1,0,,0 (0),1,1,0.0239,1 coenzyme A biosynthesis,HumanCyc,0,,7 (7),1,1,2,C00020;C00097,13 (13),0.00465,0.0577,0.0296,0.114 Coenzyme A biosynthesis,Reactome,0,,8 (9),1,1,2,C00020;C00097,13 (13),0.00465,0.0577,0.0296,0.114 Serine biosynthesis,Reactome,0,,3 (3),1,1,2,C00197;C00025,13 (13),0.00465,0.0577,0.0296,0.114 MHC class II antigen presentation,Reactome,2,SEC24C;SEC13,59 (65),0.00526,1,0,,3 (3),1,1,0.0329,0.124 Histidine metabolism,EHMN,0,,35 (36),1,1,3,C00020;C00025;C00135,45 (45),0.00539,0.0652,0.0335,0.124 Taste transduction - Homo sapiens (human),KEGG,0,,49 (53),1,1,2,C00130;C00025,13 (13),0.00465,0.0577,0.0296,0.114 Conjugation of carboxylic acids,Reactome,0,,8 (9),1,1,2,C00020;C00180,14 (15),0.0054,0.0652,0.0336,0.124 Amino Acid conjugation,Reactome,0,,8 (9),1,1,2,C00020;C00180,14 (15),0.0054,0.0652,0.0336,0.124 citrulline-nitric oxide cycle,HumanCyc,0,,5 (5),1,1,2,C00327;C00020,14 (14),0.0054,0.0652,0.0336,0.124 proline biosynthesis,HumanCyc,0,,4 (4),1,1,2,C00148;C00025,14 (14),0.0054,0.0652,0.0336,0.124 Histidine catabolism,Reactome,0,,5 (5),1,1,2,C00025;C00135,14 (14),0.0054,0.0652,0.0336,0.124 histidine degradation,HumanCyc,0,,6 (6),1,1,2,C00025;C00135,14 (14),0.0054,0.0652,0.0336,0.124 NAD biosynthesis from 2-amino-3-carboxymuconate semialdehyde,HumanCyc,0,,5 (6),1,1,2,C00020;C00025,14 (14),0.0054,0.0652,0.0336,0.124 Disulfiram Action Pathway,SMPDB,1,ACSS2,25 (25),0.0454,1,2,C00020;C00025,75 (77),0.122,0.566,0.0343,0.126 Proximal tubule bicarbonate reclamation - Homo sapiens (human),KEGG,1,GLS,23 (23),0.0419,1,1,C00025,17 (17),0.129,0.59,0.0336,0.124 The citric acid (TCA) cycle and respiratory electron transport,Reactome,0,,147 (156),1,1,3,C00158;C00020;C00989,47 (49),0.00609,0.0729,0.0371,0.136 Bile salt and organic anion SLC transporters,Reactome,0,,13 (14),1,1,2,C00158;C00791,15 (17),0.0062,0.0731,0.0377,0.136 Creatine metabolism,Reactome,0,,11 (11),1,1,2,C00077;C00791,15 (15),0.0062,0.0731,0.0377,0.136 ornithine de novo biosynthesis,HumanCyc,0,,5 (5),1,1,2,C00077;C00025,15 (15),0.0062,0.0731,0.0377,0.136 Vitamin B5 (pantothenate) metabolism,Reactome,0,,12 (13),1,1,2,C00020;C00097,15 (15),0.0062,0.0731,0.0377,0.136 Peptide hormone metabolism,Reactome,0,,49 (50),1,1,2,C01571;C06423,15 (16),0.0062,0.0731,0.0377,0.136 glycolysis,HumanCyc,1,PGK2,25 (26),0.0454,1,1,C00197,19 (20),0.143,0.636,0.0392,0.14 Phosphatidylinositol phosphate metabolism,EHMN,0,,92 (95),1,1,3,C00116;C03546;C00346,49 (49),0.00684,0.0804,0.0409,0.146 Abacavir metabolism,Reactome,0,,5 (5),1,1,2,C00020;C00130,16 (23),0.00705,0.0823,0.042,0.149 lysine degradation I (saccharopine pathway),HumanCyc,0,,4 (4),1,1,2,C00025;C00047,16 (16),0.00705,0.0823,0.042,0.149 Valine_ leucine and isoleucine degradation,EHMN,1,IVD,49 (49),0.0872,1,2,C00020;C00025,62 (62),0.0884,0.466,0.0452,0.16 Branched-chain amino acid catabolism,Reactome,1,IVD,17 (17),0.0311,1,1,C00025,36 (36),0.254,0.98,0.0461,0.162 Abacavir transport and metabolism,Reactome,0,,10 (10),1,1,2,C00020;C00130,17 (24),0.00795,0.0916,0.0464,0.162 Biogenic Amine Synthesis,Wikipathways,0,,15 (15),1,1,2,C00025;C00135,17 (17),0.00795,0.0916,0.0464,0.162 4-hydroxybenzoate biosynthesis,HumanCyc,0,,1 (1),1,1,2,C00020;C00025,17 (17),0.00795,0.0916,0.0464,0.162 Transport of glycerol from adipocytes to the liver by Aquaporins,Reactome,0,,2 (5),1,1,1,C00116,1 (1),0.00806,0.0921,0.0469,0.163 Lysine metabolism,EHMN,0,,61 (62),1,1,3,C00020;C00047;C00025,53 (53),0.00851,0.0968,0.0491,0.17 gluconeogenesis,HumanCyc,1,PGK2,26 (27),0.0472,1,1,C00197,25 (26),0.184,0.767,0.0499,0.171 Gamma-glutamyl-transpeptidase deficiency,SMPDB,0,,11 (11),1,1,2,C00025;C00097,18 (18),0.0089,0.0984,0.0509,0.171 5-oxoprolinase deficiency,SMPDB,0,,11 (11),1,1,2,C00025;C00097,18 (18),0.0089,0.0984,0.0509,0.171 Gamma-Glutamyltransferase Deficiency,SMPDB,0,,11 (11),1,1,2,C00025;C00097,18 (18),0.0089,0.0984,0.0509,0.171 Glutathione Metabolism,SMPDB,0,,11 (11),1,1,2,C00025;C00097,18 (18),0.0089,0.0984,0.0509,0.171 Glutathione Synthetase Deficiency,SMPDB,0,,11 (11),1,1,2,C00025;C00097,18 (18),0.0089,0.0984,0.0509,0.171 5-Oxoprolinuria,SMPDB,0,,11 (11),1,1,2,C00025;C00097,18 (18),0.0089,0.0984,0.0509,0.171 serine and glycine biosynthesis,HumanCyc,0,,8 (8),1,1,2,C00197;C00025,18 (18),0.0089,0.0984,0.0509,0.171 Glutathione metabolism,Wikipathways,0,,20 (20),1,1,2,C00025;C00097,18 (18),0.0089,0.0984,0.0509,0.171 Glycosphingolipid metabolism,EHMN,0,,69 (69),1,1,3,C00031;C00836;C00346,55 (55),0.00943,0.104,0.0534,0.175 Tyrosine metabolism,EHMN,0,,105 (110),1,1,4,C00020;C00041;C00025;C00097,105 (105),0.00945,0.104,0.0535,0.175 Purine metabolism - Homo sapiens (human),KEGG,0,,170 (176),1,1,4,C00385;C00020;C00130;C00366,92 (92),0.00595,0.0715,0.0364,0.134 Fatty Acid Biosynthesis,SMPDB,0,,2 (2),1,1,2,C01571;C06423,19 (31),0.0099,0.108,0.0556,0.175 Pyrimidine metabolism,Reactome,0,,25 (25),1,1,3,C00106;C00041;C00025,56 (57),0.00991,0.108,0.0556,0.175 Warburg Effect,SMPDB,0,,45 (45),1,1,3,C00158;C00031;C00025,56 (56),0.00991,0.108,0.0556,0.175 Glycolysis and Gluconeogenesis,Wikipathways,1,PGK2,49 (49),0.0872,1,1,C00031,15 (16),0.115,0.54,0.056,0.175 Defective ALG2 causes ALG2-CDG (CDG-1i),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective DPAGT1 causes DPAGT1-CDG (CDG-1j) and CMSTA2,Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective ALG1 causes ALG1-CDG (CDG-1k),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective ALG11 causes ALG11-CDG (CDG-1p),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Asparagine N-linked glycosylation,Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective RFT1 causes RFT1-CDG (CDG-1n),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective MPDU1 causes MPDU1-CDG (CDG-1f),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective ALG8 causes ALG8-CDG (CDG-1h),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective ALG6 causes ALG6-CDG (CDG-1c),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective ALG14 causes congenital myasthenic syndrome (ALG14-CMS),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Diseases associated with N-glycosylation of proteins,Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective ALG9 causes ALG9-CDG (CDG-1l),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective ALG3 causes ALG3-CDG (CDG-1d),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective B4GALT1 causes B4GALT1-CDG (CDG-2d),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective ALG12 causes ALG12-CDG (CDG-1g),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective MAN1B1 causes MRT15,Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective MGAT2 causes MGAT2-CDG (CDG-2a),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Defective MOGS causes MOGS-CDG (CDG-2b),Reactome,2,SEC24C;SEC13,119 (122),0.0202,1,1,C00025,84 (94),0.497,1,0.0562,0.175 Retrograde endocannabinoid signaling - Homo sapiens (human),KEGG,0,,94 (103),1,1,2,C00116;C00025,19 (19),0.0099,0.108,0.0556,0.175 GABAergic synapse - Homo sapiens (human),KEGG,1,GLS,84 (90),0.145,1,1,C00025,9 (9),0.0703,0.401,0.057,0.177 Cysteine and methionine metabolism - Homo sapiens (human),KEGG,0,,37 (38),1,1,3,C00041;C00097;C00170,57 (57),0.0104,0.113,0.0579,0.179 Nonhomologous End-joining (NHEJ),Reactome,1,XRCC6,6 (6),0.0111,1,0,,2 (2),1,1,0.061,0.187 Glyoxylate and dicarboxylate metabolism - Homo sapiens (human),KEGG,0,,25 (25),1,1,3,C00158;C00197;C00025,58 (58),0.0109,0.118,0.0602,0.185 lysine degradation II (pipecolate pathway),HumanCyc,0,,6 (6),1,1,2,C00025;C00047,21 (21),0.012,0.128,0.0652,0.197 Pantothenate and CoA Biosynthesis,SMPDB,0,,6 (6),1,1,2,C00020;C00097,21 (21),0.012,0.128,0.0652,0.197 Lysine catabolism,Reactome,0,,8 (8),1,1,2,C00047;C00025,21 (21),0.012,0.128,0.0652,0.197 Vitamin B5 - CoA biosynthesis from pantothenate,EHMN,0,,31 (32),1,1,2,C00020;C00097,21 (21),0.012,0.128,0.0652,0.197 Pyruvate Dehydrogenase Complex Deficiency,SMPDB,1,ACSS2,22 (22),0.0401,1,1,C00020,46 (46),0.312,1,0.0674,0.2 Primary hyperoxaluria II_ PH2,SMPDB,1,ACSS2,22 (22),0.0401,1,1,C00020,46 (46),0.312,1,0.0674,0.2 Pyruvate kinase deficiency,SMPDB,1,ACSS2,22 (22),0.0401,1,1,C00020,46 (46),0.312,1,0.0674,0.2 Leigh Syndrome,SMPDB,1,ACSS2,22 (22),0.0401,1,1,C00020,46 (46),0.312,1,0.0674,0.2 Pyruvate Metabolism,SMPDB,1,ACSS2,22 (22),0.0401,1,1,C00020,46 (46),0.312,1,0.0674,0.2 Pyruvate Decarboxylase E1 Component Deficiency (PDHE1 Deficiency),SMPDB,1,ACSS2,22 (22),0.0401,1,1,C00020,46 (46),0.312,1,0.0674,0.2 Fatty Acid Beta Oxidation,Wikipathways,1,ACSS2,34 (34),0.0613,1,1,C00116,29 (32),0.21,0.847,0.0689,0.204 Hemostasis,Reactome,0,,471 (501),1,1,3,C00020;C00327;C00116,62 (63),0.0131,0.139,0.0698,0.206 Transfer of Acetyl Groups into Mitochondria,SMPDB,0,,9 (9),1,1,2,C00158;C00031,22 (22),0.0132,0.139,0.0702,0.207 Iron uptake and transport,Wikipathways,0,,0 (0),1,1,2,C00020;C00366,29 (31),0.0224,0.215,0.0224,0.215 Passive Transport by Aquaporins,Wikipathways,0,,0 (0),1,1,1,C00116,3 (3),0.024,0.219,0.024,0.219 Rilpivirine Action Pathway,SMPDB,0,,0 (0),1,1,1,C00020,3 (3),0.024,0.219,0.024,0.219 ISG15 antiviral mechanism,Wikipathways,0,,0 (0),1,1,1,C00020,3 (3),0.024,0.219,0.024,0.219 SUMOylation,Wikipathways,0,,0 (0),1,1,1,C00020,3 (3),0.024,0.219,0.024,0.219 Glutamatergic synapse - Homo sapiens (human),KEGG,1,GLS,106 (116),0.18,1,1,C00025,8 (8),0.0628,0.379,0.0618,0.189 Ion channel transport,Reactome,0,,162 (173),1,1,2,C00020;C00366,23 (24),0.0144,0.151,0.0753,0.221 Recycling of bile acids and salts,Reactome,0,,15 (16),1,1,2,C00020;C00245,24 (24),0.0156,0.163,0.0804,0.234 Phenylalanine and tyrosine catabolism,Reactome,0,,9 (10),1,1,2,C00041;C00025,24 (24),0.0156,0.163,0.0804,0.234 AMPK signaling pathway - Homo sapiens (human),KEGG,0,,121 (124),1,1,2,C00020;C00031,23 (23),0.0144,0.151,0.0753,0.221 Deadenylation of mRNA,Reactome,0,,31 (32),1,1,1,C00020,2 (3),0.0161,0.166,0.0824,0.238 Presynaptic function of Kainate receptors,Reactome,0,,19 (21),1,1,1,C00025,2 (2),0.0161,0.166,0.0824,0.238 basic mechanisms of sumoylation,BioCarta,0,,6 (7),1,1,1,C00020,2 (2),0.0161,0.166,0.0824,0.238 Fatty acid_ triacylglycerol_ and ketone body metabolism,Wikipathways,0,,0 (0),1,1,3,C00158;C00020;C00116,81 (89),0.0266,0.239,0.0266,0.239 Non-alcoholic fatty liver disease (NAFLD) - Homo sapiens (human),KEGG,0,,146 (151),1,1,1,C00031,2 (2),0.0161,0.166,0.0824,0.238 Lysine Degradation,SMPDB,0,,13 (13),1,1,2,C00047;C00025,25 (27),0.0169,0.169,0.0857,0.24 Hyperlysinemia I_ Familial,SMPDB,0,,13 (13),1,1,2,C00047;C00025,25 (27),0.0169,0.169,0.0857,0.24 2-aminoadipic 2-oxoadipic aciduria,SMPDB,0,,13 (13),1,1,2,C00047;C00025,25 (27),0.0169,0.169,0.0857,0.24 Pyridoxine dependency with seizures,SMPDB,0,,13 (13),1,1,2,C00047;C00025,25 (27),0.0169,0.169,0.0857,0.24 Saccharopinuria/Hyperlysinemia II,SMPDB,0,,13 (13),1,1,2,C00047;C00025,25 (27),0.0169,0.169,0.0857,0.24 Hyperlysinemia II or Saccharopinuria,SMPDB,0,,13 (13),1,1,2,C00047;C00025,25 (27),0.0169,0.169,0.0857,0.24 methionine salvage cycle III,HumanCyc,0,,10 (10),1,1,2,C00170;C00315,25 (25),0.0169,0.169,0.0857,0.24 Tryptophan catabolism,Reactome,0,,11 (13),1,1,2,C00041;C00025,25 (26),0.0169,0.169,0.0857,0.24 Glutaric Aciduria Type III,SMPDB,0,,13 (13),1,1,2,C00047;C00025,25 (27),0.0169,0.169,0.0857,0.24 Neurotransmitter Release Cycle,Reactome,1,GLS,43 (44),0.0769,1,1,C00025,31 (31),0.223,0.892,0.0868,0.243 Organic cation transport,Reactome,0,,8 (8),1,1,2,C00315;C00791,26 (29),0.0182,0.181,0.091,0.253 Degradation of cysteine and homocysteine,Reactome,0,,8 (8),1,1,2,C00245;C00097,26 (29),0.0182,0.181,0.091,0.253 Glycerophospholipid Biosynthetic Pathway,Wikipathways,0,,0 (0),1,1,2,C00116;C00346,35 (51),0.0318,0.26,0.0318,0.26 Methylhistidine Metabolism,SMPDB,0,,0 (0),1,1,1,C00135,4 (4),0.0319,0.26,0.0319,0.26 Zidovudine Action Pathway,SMPDB,0,,0 (0),1,1,1,C00020,4 (4),0.0319,0.26,0.0319,0.26 Zalcitabine Action Pathway,SMPDB,0,,0 (0),1,1,1,C00020,4 (4),0.0319,0.26,0.0319,0.26 Stavudine Action Pathway,SMPDB,0,,0 (0),1,1,1,C00020,4 (4),0.0319,0.26,0.0319,0.26 Nevirapine Action Pathway,SMPDB,0,,0 (0),1,1,1,C00020,4 (4),0.0319,0.26,0.0319,0.26 Lamivudine Action Pathway,SMPDB,0,,0 (0),1,1,1,C00020,4 (4),0.0319,0.26,0.0319,0.26 Emtricitabine Action Pathway,SMPDB,0,,0 (0),1,1,1,C00020,4 (4),0.0319,0.26,0.0319,0.26 Efavirenz Action Pathway,SMPDB,0,,0 (0),1,1,1,C00020,4 (4),0.0319,0.26,0.0319,0.26 Abacavir Action Pathway,SMPDB,0,,0 (0),1,1,1,C00020,4 (4),0.0319,0.26,0.0319,0.26 Didanosine Action Pathway,SMPDB,0,,0 (0),1,1,1,C00020,4 (4),0.0319,0.26,0.0319,0.26 Delavirdine Action Pathway,SMPDB,0,,0 (0),1,1,1,C00020,4 (4),0.0319,0.26,0.0319,0.26 Sphingolipid metabolism - Homo sapiens (human),KEGG,0,,46 (47),1,1,2,C00836;C00346,25 (25),0.0169,0.169,0.0857,0.24 Fatty Acyl-CoA Biosynthesis,Reactome,0,,21 (22),1,1,2,C00158;C00020,27 (30),0.0195,0.193,0.0964,0.264 Nicotinate metabolism,Reactome,0,,11 (12),1,1,2,C00020;C00025,27 (27),0.0195,0.193,0.0964,0.264 3-Methylglutaconic Aciduria Type I,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 Valine_ Leucine and Isoleucine Degradation,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 2-Methyl-3-Hydroxybutryl CoA Dehydrogenase Deficiency,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 Isovaleric Aciduria,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 3-Methylcrotonyl Coa Carboxylase Deficiency Type I,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 Propionic Acidemia,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 Maple Syrup Urine Disease,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 3-Hydroxy-3-Methylglutaryl-CoA Lyase Deficiency,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 Isobutyryl-coa dehydrogenase deficiency,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 3-hydroxyisobutyric aciduria,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 3-hydroxyisobutyric acid dehydrogenase deficiency,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 Isovaleric acidemia,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 Methylmalonate Semialdehyde Dehydrogenase Deficiency,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 Methylmalonic Aciduria,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 3-Methylglutaconic Aciduria Type IV,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 3-Methylglutaconic Aciduria Type III,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 Beta-Ketothiolase Deficiency,SMPDB,1,IVD,30 (30),0.0543,1,1,C00025,58 (58),0.377,1,0.1,0.264 Phenylalanine and Tyrosine Metabolism,SMPDB,0,,11 (11),1,1,2,C00020;C00025,28 (28),0.0209,0.202,0.102,0.264 Phenylketonuria,SMPDB,0,,11 (11),1,1,2,C00020;C00025,28 (28),0.0209,0.202,0.102,0.264 Selenoamino Acid Metabolism,SMPDB,0,,11 (12),1,1,2,C00020;C00041,28 (28),0.0209,0.202,0.102,0.264 Tyrosinemia Type 3 (TYRO3),SMPDB,0,,11 (11),1,1,2,C00020;C00025,28 (28),0.0209,0.202,0.102,0.264 Tyrosinemia Type 2 (or Richner-Hanhart syndrome),SMPDB,0,,11 (11),1,1,2,C00020;C00025,28 (28),0.0209,0.202,0.102,0.264 Carbohydrate digestion and absorption - Homo sapiens (human),KEGG,0,,45 (45),1,1,2,C00031;C01835,27 (27),0.0195,0.193,0.0964,0.264 Glycerolipid metabolism - Homo sapiens (human),KEGG,0,,55 (55),1,1,2,C00116;C00197,28 (28),0.0209,0.202,0.102,0.264 Pantothenate and CoA biosynthesis - Homo sapiens (human),KEGG,0,,17 (17),1,1,2,C00106;C00097,28 (28),0.0209,0.202,0.102,0.264 Pyrimidine metabolism,EHMN,0,,134 (139),1,1,3,C00020;C00106;C00025,77 (77),0.0233,0.219,0.111,0.282 Metabolism of lipids and lipoproteins,Reactome,0,,497 (516),1,1,7,C00158;C00116;C00245;C00020;C00025;C00836;C00346,365 (447),0.0237,0.219,0.112,0.282 L-kynurenine degradation,HumanCyc,0,,13 (15),1,1,2,C00041;C00025,30 (30),0.0238,0.219,0.113,0.282 DNA-PK pathway in nonhomologous end joining,PID,1,XRCC6,13 (13),0.0239,1,0,,1 (1),1,1,0.113,0.282 Antiviral mechanism by IFN-stimulated genes,Reactome,0,,40 (40),1,1,1,C00020,3 (4),0.024,0.219,0.114,0.282 Cleavage of the damaged pyrimidine ,Reactome,0,,6 (6),1,1,1,C00106,3 (8),0.024,0.219,0.114,0.282 Sulfur metabolism - Homo sapiens (human),KEGG,0,,9 (10),1,1,2,C00245;C00097,29 (29),0.0224,0.215,0.107,0.278 Activation of Ca-permeable Kainate Receptor,Reactome,0,,12 (12),1,1,1,C00025,3 (3),0.024,0.219,0.114,0.282 AMPK inhibits chREBP transcriptional activation activity,Reactome,0,,5 (5),1,1,1,C00020,3 (3),0.024,0.219,0.114,0.282 Activation of Na-permeable Kainate Receptors,Reactome,0,,2 (2),1,1,1,C00025,3 (3),0.024,0.219,0.114,0.282 Activation of PPARGC1A (PGC-1alpha) by phosphorylation,Reactome,0,,10 (10),1,1,1,C00020,3 (3),0.024,0.219,0.114,0.282 Processing and activation of SUMO,Reactome,0,,9 (10),1,1,1,C00020,3 (3),0.024,0.219,0.114,0.282 Activation of AMPA receptors,Reactome,0,,2 (4),1,1,1,C00025,3 (3),0.024,0.219,0.114,0.282 Depyrimidination,Reactome,0,,6 (6),1,1,1,C00106,3 (8),0.024,0.219,0.114,0.282 SUMO is conjugated to E1 (UBA2:SAE1),Reactome,0,,5 (6),1,1,1,C00020,3 (3),0.024,0.219,0.114,0.282 ISG15 antiviral mechanism,Reactome,0,,40 (40),1,1,1,C00020,3 (4),0.024,0.219,0.114,0.282 SUMOylation,Reactome,0,,9 (10),1,1,1,C00020,3 (3),0.024,0.219,0.114,0.282 Tryptophan metabolism,EHMN,0,,22 (22),1,1,3,C00020;C00041;C00025,78 (78),0.0241,0.219,0.114,0.282 The citric acid (TCA) cycle and respiratory electron transport,Wikipathways,0,,0 (0),1,1,2,C00158;C00989,37 (37),0.0352,0.283,0.0352,0.283 Sphingolipid de novo biosynthesis,Reactome,0,,32 (33),1,1,2,C00836;C00346,31 (31),0.0253,0.229,0.119,0.292 Purine salvage,Reactome,0,,13 (14),1,1,2,C00020;C00130,31 (32),0.0253,0.229,0.119,0.292 IRF3-mediated induction of type I IFN,Reactome,1,XRCC6,14 (14),0.0257,1,0,,5 (7),1,1,0.12,0.293 telomeres telomerase cellular aging and immortality,BioCarta,1,XRCC6,14 (15),0.0257,1,0,,3 (3),1,1,0.12,0.293 Early Phase of HIV Life Cycle,Reactome,1,XRCC6,14 (16),0.0257,1,0,,5 (5),1,1,0.12,0.293 Amino acid conjugation,Wikipathways,0,,0 (0),1,1,1,C00245,5 (5),0.0397,0.3,0.0397,0.3 Eukaryotic Translation Elongation,Reactome,2,EEF1D;RPL9,138 (148),0.0266,1,0,,24 (25),1,1,0.123,0.301 Huntington_s disease - Homo sapiens (human),KEGG,0,,190 (193),1,1,1,C00025,3 (3),0.024,0.219,0.114,0.282 Propanoate metabolism,EHMN,0,,18 (18),1,1,2,C00158;C00020,33 (33),0.0285,0.255,0.13,0.316 Fatty acid_ triacylglycerol_ and ketone body metabolism,Reactome,0,,94 (96),1,1,3,C00158;C00020;C00116,85 (91),0.0301,0.26,0.136,0.325 STING mediated induction of host immune responses,Reactome,1,XRCC6,17 (17),0.0311,1,0,,7 (9),1,1,0.139,0.325 Nicotinate and Nicotinamide Metabolism,SMPDB,0,,14 (14),1,1,2,C00020;C00025,35 (35),0.0318,0.26,0.141,0.325 putrescine biosynthesis I,HumanCyc,0,,1 (1),1,1,1,C00077,4 (4),0.0319,0.26,0.142,0.325 Mitochondrial biogenesis,Reactome,0,,26 (28),1,1,1,C00020,4 (4),0.0319,0.26,0.142,0.325 Butirosin and neomycin biosynthesis - Homo sapiens (human),KEGG,0,,5 (5),1,1,2,C00031;C00025,33 (33),0.0285,0.255,0.13,0.316 Pancreas Function,SMPDB,0,,6 (6),1,1,1,C00031,4 (4),0.0319,0.26,0.142,0.325 hydrogen sulfide biosynthesis (trans-sulfuration),HumanCyc,0,,2 (3),1,1,1,C00097,4 (6),0.0319,0.26,0.142,0.325 Processive synthesis on the lagging strand,Reactome,0,,16 (16),1,1,1,C00020,4 (4),0.0319,0.26,0.142,0.325 S-methyl-5_-thioadenosine degradation,HumanCyc,0,,1 (1),1,1,1,C00170,4 (4),0.0319,0.26,0.142,0.325 Ionotropic activity of Kainate Receptors,Reactome,0,,12 (12),1,1,1,C00025,4 (4),0.0319,0.26,0.142,0.325 glutamate dependent acid resistance,HumanCyc,0,,3 (3),1,1,1,C00025,4 (4),0.0319,0.26,0.142,0.325 Ribosome biogenesis in eukaryotes - Homo sapiens (human),KEGG,1,REXO2,81 (84),0.14,1,0,,0 (0),1,1,0.14,1 aspartate biosynthesis,HumanCyc,0,,2 (2),1,1,1,C00025,4 (4),0.0319,0.26,0.142,0.325 Activation of Kainate Receptors upon glutamate binding,Reactome,0,,30 (32),1,1,1,C00025,4 (4),0.0319,0.26,0.142,0.325 Antigen processing: Ubiquitination & Proteasome degradation,Reactome,0,,163 (178),1,1,1,C00020,4 (4),0.0319,0.26,0.142,0.325 Removal of the Flap Intermediate,Reactome,0,,14 (14),1,1,1,C00020,4 (4),0.0319,0.26,0.142,0.325 histamine biosynthesis,HumanCyc,0,,1 (1),1,1,1,C00135,4 (4),0.0319,0.26,0.142,0.325 Type II diabetes mellitus,Wikipathways,0,,21 (21),1,1,1,C00031,4 (4),0.0319,0.26,0.142,0.325 Sodium-coupled sulphate_ di- and tri-carboxylate transporters,Reactome,0,,5 (6),1,1,1,C00158,4 (4),0.0319,0.26,0.142,0.325 TCA Cycle Nutrient Utilization and Invasiveness of Ovarian Cancer,Wikipathways,0,,5 (5),1,1,1,C00031,4 (4),0.0319,0.26,0.142,0.325 Class II GLUTs,Reactome,0,,4 (4),1,1,1,C00366,4 (4),0.0319,0.26,0.142,0.325 regulators of bone mineralization,BioCarta,0,,11 (11),1,1,1,C00020,4 (4),0.0319,0.26,0.142,0.325 no2-dependent il-12 pathway in nk cells,BioCarta,0,,8 (9),1,1,1,C00327,4 (4),0.0319,0.26,0.142,0.325 Cam-PDE 1 activation,Reactome,0,,4 (6),1,1,1,C00020,4 (4),0.0319,0.26,0.142,0.325 glycine biosynthesis,HumanCyc,0,,2 (2),1,1,1,C00041,4 (4),0.0319,0.26,0.142,0.325 superpathway of methionine degradation,HumanCyc,0,,19 (20),1,1,2,C00025;C00097,36 (36),0.0335,0.27,0.147,0.326 Butanoate metabolism,EHMN,0,,24 (24),1,1,2,C00020;C00025,36 (36),0.0335,0.27,0.147,0.326 Vitamin B3 (nicotinate and nicotinamide) metabolism,EHMN,0,,47 (47),1,1,2,C00020;C00025,36 (36),0.0335,0.27,0.147,0.326 Galactose Metabolism,SMPDB,0,,13 (13),1,1,2,C00116;C00031,36 (36),0.0335,0.27,0.147,0.326 Galactosemia,SMPDB,0,,13 (13),1,1,2,C00116;C00031,36 (36),0.0335,0.27,0.147,0.326 Phase 1 - Functionalization of compounds,Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP27A1 causes Cerebrotendinous xanthomatosis (CTX),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP1B1 causes Glaucoma,Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP4F22 causes Ichthyosis_ congenital_ autosomal recessive 5 (ARCI5),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP11B2 causes Corticosterone methyloxidase 1 deficiency (CMO-1 deficiency),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP24A1 causes Hypercalcemia_ infantile (HCAI),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective MAOA causes Brunner syndrome (BRUNS),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP7B1 causes Spastic paraplegia 5A_ autosomal recessive (SPG5A) and Congenital bile acid synthesis defect 3 (CBAS3),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective FMO3 causes Trimethylaminuria (TMAU),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP21A2 causes Adrenal hyperplasia 3 (AH3),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP11A1 causes Adrenal insufficiency_ congenital_ with 46_XY sex reversal (AICSR),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP2R1 causes Rickets vitamin D-dependent 1B (VDDR1B),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP17A1 causes Adrenal hyperplasia 5 (AH5),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP11B1 causes Adrenal hyperplasia 4 (AH4),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP19A1 causes Aromatase excess syndrome (AEXS),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP26B1 causes Radiohumeral fusions with other skeletal and craniofacial anomalies (RHFCA),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP26C1 causes Focal facial dermal dysplasia 4 (FFDD4),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP27B1 causes Rickets vitamin D-dependent 1A (VDDR1A),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective TBXAS1 causes Ghosal hematodiaphyseal dysplasia (GHDD),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Defective CYP2U1 causes Spastic paraplegia 56_ autosomal recessive (SPG56),Reactome,1,ACSS2,73 (80),0.127,1,2,C00020;C00315,126 (151),0.27,1,0.15,0.326 Pentose phosphate pathway,EHMN,0,,29 (29),1,1,2,C00020;C00257,37 (37),0.0352,0.283,0.153,0.332 Glycerophospholipid Biosynthetic Pathway,Wikipathways,0,,0 (0),1,1,1,C00116,6 (9),0.0475,0.336,0.0475,0.336 Translocation of GLUT4 to the Plasma Membrane,Wikipathways,0,,0 (0),1,1,1,C00020,6 (6),0.0475,0.336,0.0475,0.336 Neurotransmitter uptake and Metabolism In Glial Cells,Wikipathways,0,,0 (0),1,1,1,C00025,6 (6),0.0475,0.336,0.0475,0.336 Deadenylation-dependent mRNA decay,Wikipathways,0,,0 (0),1,1,1,C00020,6 (7),0.0475,0.336,0.0475,0.336 Malonyl-coa decarboxylase deficiency,SMPDB,0,,14 (14),1,1,2,C00020;C00025,38 (38),0.037,0.292,0.159,0.339 Malonic Aciduria,SMPDB,0,,14 (14),1,1,2,C00020;C00025,38 (38),0.037,0.292,0.159,0.339 Propanoate Metabolism,SMPDB,0,,14 (14),1,1,2,C00020;C00025,38 (38),0.037,0.292,0.159,0.339 Methylmalonic Aciduria Due to Cobalamin-Related Disorders,SMPDB,0,,14 (14),1,1,2,C00020;C00025,38 (38),0.037,0.292,0.159,0.339 Selenoamino acid metabolism,EHMN,0,,35 (36),1,1,2,C00020;C00041,38 (38),0.037,0.292,0.159,0.339 Pyrimidine catabolism,Reactome,0,,12 (12),1,1,2,C00106;C00041,38 (39),0.037,0.292,0.159,0.339 Pyruvate metabolism and Citric Acid (TCA) cycle,Reactome,0,,44 (47),1,1,2,C00158;C00989,38 (38),0.037,0.292,0.159,0.339 PI3K-Akt signaling pathway - Homo sapiens (human),KEGG,0,,335 (347),1,1,1,C00020,4 (4),0.0319,0.26,0.142,0.325 Valproic Acid Pathway_ Pharmacokinetics,PharmGKB,1,IVD,21 (21),0.0383,1,0,,1 (1),1,1,0.163,0.343 Diseases of glycosylation,Reactome,2,SEC24C;SEC13,226 (230),0.065,1,1,C00025,109 (139),0.591,1,0.164,0.343 Synthesis and interconversion of nucleotide di- and triphosphates,Reactome,0,,19 (19),1,1,2,C00020;C00025,39 (39),0.0388,0.3,0.165,0.343 GABA synthesis,Reactome,0,,2 (2),1,1,1,C00025,5 (5),0.0397,0.3,0.168,0.343 Facilitative Na+-independent glucose transporters,Reactome,0,,11 (12),1,1,1,C00366,5 (5),0.0397,0.3,0.168,0.343 Regulation of AMPK activity via LKB1,Reactome,0,,14 (14),1,1,1,C00020,5 (5),0.0397,0.3,0.168,0.343 Glibenclamide Action Pathway,SMPDB,0,,6 (6),1,1,1,C00031,5 (5),0.0397,0.3,0.168,0.343 Gliclazide Action Pathway,SMPDB,0,,6 (6),1,1,1,C00031,5 (5),0.0397,0.3,0.168,0.343 Repaglinide Action Pathway,SMPDB,0,,6 (6),1,1,1,C00031,5 (5),0.0397,0.3,0.168,0.343 Nateglinide Action Pathway,SMPDB,0,,6 (6),1,1,1,C00031,5 (5),0.0397,0.3,0.168,0.343 mRNA decay by 5_ to 3_ exoribonuclease,Reactome,0,,13 (15),1,1,1,C00020,5 (7),0.0397,0.3,0.168,0.343 hypusine biosynthesis,HumanCyc,0,,2 (2),1,1,1,C00315,5 (5),0.0397,0.3,0.168,0.343 PRPP biosynthesis,HumanCyc,0,,3 (3),1,1,1,C00020,5 (6),0.0397,0.3,0.168,0.343 adenine and adenosine salvage II,HumanCyc,0,,1 (1),1,1,1,C00020,5 (5),0.0397,0.3,0.168,0.343 thiamin salvage III,HumanCyc,0,,1 (1),1,1,1,C00020,5 (5),0.0397,0.3,0.168,0.343 Removal of the Flap Intermediate from the C-strand,Reactome,0,,10 (10),1,1,1,C00020,5 (5),0.0397,0.3,0.168,0.343 Ras activation uopn Ca2+ infux through NMDA receptor,Reactome,0,,18 (19),1,1,1,C00025,5 (5),0.0397,0.3,0.168,0.343 Unblocking of NMDA receptor_ glutamate binding and activation,Reactome,0,,14 (17),1,1,1,C00025,5 (5),0.0397,0.3,0.168,0.343 CREB phosphorylation through the activation of CaMKII,Reactome,0,,16 (17),1,1,1,C00025,5 (5),0.0397,0.3,0.168,0.343 hypoxia-inducible factor in the cardivascular system,BioCarta,0,,15 (16),1,1,1,C00031,5 (5),0.0397,0.3,0.168,0.343 nitric oxide signaling pathway,BioCarta,0,,18 (19),1,1,1,C00327,5 (5),0.0397,0.3,0.168,0.343 Fatty Acid Biosynthesis,Wikipathways,1,ACSS2,22 (22),0.0401,1,0,,4 (4),1,1,0.169,0.344 Glycerophospholipid metabolism,EHMN,0,,137 (141),1,1,3,C00116;C00041;C00346,96 (96),0.0411,0.31,0.172,0.35 Double-Strand Break Repair,Reactome,1,XRCC6,23 (23),0.0419,1,0,,2 (2),1,1,0.175,0.355 Insulin signaling pathway - Homo sapiens (human),KEGG,0,,136 (141),1,1,1,C00031,4 (4),0.0319,0.26,0.142,0.325 GPCR ligand binding,Wikipathways,0,,0 (0),1,1,3,C00077;C00047;C00025,106 (125),0.0526,0.361,0.0526,0.361 Energy dependent regulation of mTOR by LKB1-AMPK,Wikipathways,0,,0 (0),1,1,1,C00020,7 (7),0.0551,0.361,0.0551,0.361 Sphingolipid metabolism,Wikipathways,0,,0 (0),1,1,1,C00836,7 (7),0.0551,0.361,0.0551,0.361 bile acid biosynthesis_ neutral pathway,HumanCyc,0,,15 (15),1,1,2,C00020;C00245,42 (44),0.0444,0.334,0.183,0.369 Aminosugars metabolism,EHMN,0,,50 (51),1,1,2,C00041;C00025,43 (43),0.0464,0.336,0.189,0.372 chondroitin sulfate degradation (metazoa),HumanCyc,0,,6 (7),1,1,3,C00379;C00116;C01835,101 (137),0.0466,0.336,0.19,0.372 D-myo-inositol (1_4_5)-trisphosphate degradation,HumanCyc,0,,13 (13),1,1,1,C03546,6 (6),0.0475,0.336,0.192,0.372 Astrocytic Glutamate-Glutamine Uptake And Metabolism,Reactome,0,,4 (4),1,1,1,C00025,6 (6),0.0475,0.336,0.192,0.372 Neurotransmitter uptake and Metabolism In Glial Cells,Reactome,0,,4 (4),1,1,1,C00025,6 (6),0.0475,0.336,0.192,0.372 α-tocopherol degradation,HumanCyc,0,,1 (1),1,1,1,C02477,6 (9),0.0475,0.336,0.192,0.372 PDE3B signalling,Reactome,0,,2 (2),1,1,1,C00020,6 (6),0.0475,0.336,0.192,0.372 Regulation of Rheb GTPase activity by AMPK,Reactome,0,,9 (10),1,1,1,C00020,6 (6),0.0475,0.336,0.192,0.372 Protein processing in endoplasmic reticulum - Homo sapiens (human),KEGG,2,SEC24C;SEC13,167 (168),0.0378,1,0,,6 (6),1,1,0.162,0.343 homocarnosine biosynthesis,HumanCyc,0,,1 (1),1,1,1,C00135,6 (7),0.0475,0.336,0.192,0.372 Inhibition of HSL,Reactome,0,,2 (2),1,1,1,C00020,6 (6),0.0475,0.336,0.192,0.372 Deadenylation-dependent mRNA decay,Reactome,0,,55 (58),1,1,1,C00020,6 (9),0.0475,0.336,0.192,0.372 Lagging Strand Synthesis,Reactome,0,,21 (21),1,1,1,C00020,6 (6),0.0475,0.336,0.192,0.372 lipoate salvage,HumanCyc,0,,2 (2),1,1,1,C00020,6 (6),0.0475,0.336,0.192,0.372 VEGFR1 specific signals,PID,0,,29 (31),1,1,1,C00327,6 (6),0.0475,0.336,0.192,0.372 Glycogen Metabolism,Wikipathways,0,,36 (36),1,1,1,C00031,6 (6),0.0475,0.336,0.192,0.372 5-Phosphoribose 1-diphosphate biosynthesis,Reactome,0,,3 (3),1,1,1,C00020,6 (6),0.0475,0.336,0.192,0.372 vegf hypoxia and angiogenesis,BioCarta,0,,27 (31),1,1,1,C00327,6 (6),0.0475,0.336,0.192,0.372 Insulin Signalling,SMPDB,0,,15 (17),1,1,1,C00031,6 (6),0.0475,0.336,0.192,0.372 Translocation of GLUT4 to the plasma membrane,Reactome,0,,34 (34),1,1,1,C00020,6 (6),0.0475,0.336,0.192,0.372 DNA strand elongation,Reactome,0,,32 (32),1,1,1,C00020,6 (6),0.0475,0.336,0.192,0.372 mRNA decay by 3_ to 5_ exoribonuclease,Reactome,0,,11 (11),1,1,1,C00020,6 (8),0.0475,0.336,0.192,0.372 Axon guidance,Reactome,1,EVL,310 (332),0.443,1,1,C00025,14 (14),0.107,0.512,0.192,0.372 dermatan sulfate degradation (metazoa),HumanCyc,0,,6 (7),1,1,3,C00379;C00116;C01835,102 (138),0.0478,0.338,0.193,0.373 Phospholipid biosynthesis,Wikipathways,0,,0 (0),1,1,1,C00346,8 (8),0.0628,0.379,0.0628,0.379 Butanoate metabolism - Homo sapiens (human),KEGG,0,,26 (26),1,1,2,C00025;C00989,41 (41),0.0425,0.32,0.177,0.358 Type II diabetes mellitus - Homo sapiens (human),KEGG,0,,45 (48),1,1,1,C00031,6 (6),0.0475,0.336,0.192,0.372 Histidine metabolism - Homo sapiens (human),KEGG,0,,23 (23),1,1,2,C00025;C00135,45 (45),0.0503,0.355,0.201,0.386 Translation,Reactome,2,EEF1D;RPL9,205 (217),0.0547,1,0,,28 (29),1,1,0.214,0.387 Rapoport-Luebering glycolytic shunt,HumanCyc,0,,4 (6),1,1,1,C00197,7 (7),0.0551,0.361,0.215,0.387 Conjugation of salicylate with glycine,Reactome,0,,7 (8),1,1,1,C00020,7 (8),0.0551,0.361,0.215,0.387 Energy dependent regulation of mTOR by LKB1-AMPK,Reactome,0,,16 (17),1,1,1,C00020,7 (7),0.0551,0.361,0.215,0.387 Galactose metabolism - Homo sapiens (human),KEGG,0,,30 (30),1,1,2,C00116;C00031,45 (45),0.0503,0.355,0.201,0.386 glycerol degradation,HumanCyc,0,,5 (5),1,1,1,C00116,7 (8),0.0551,0.361,0.215,0.387 Iminoglycinuria,SMPDB,0,,28 (33),1,1,1,C00097,7 (8),0.0551,0.361,0.215,0.387 Lysinuric Protein Intolerance,SMPDB,0,,28 (33),1,1,1,C00097,7 (8),0.0551,0.361,0.215,0.387 cysteine biosynthesis/homocysteine degradation (trans-sulfuration),HumanCyc,0,,2 (3),1,1,1,C00097,7 (7),0.0551,0.361,0.215,0.387 Blue diaper syndrome,SMPDB,0,,28 (33),1,1,1,C00097,7 (8),0.0551,0.361,0.215,0.387 Lysinuric protein intolerance (LPI),SMPDB,0,,28 (33),1,1,1,C00097,7 (8),0.0551,0.361,0.215,0.387 Iminoglycinuria,SMPDB,0,,28 (33),1,1,1,C00097,7 (8),0.0551,0.361,0.215,0.387 Cystinuria,SMPDB,0,,28 (33),1,1,1,C00097,7 (8),0.0551,0.361,0.215,0.387 Torsemide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,7 (9),0.0551,0.361,0.215,0.387 Biotin Metabolism,SMPDB,0,,3 (3),1,1,1,C00047,7 (7),0.0551,0.361,0.215,0.387 carnosine biosynthesis,HumanCyc,0,,1 (1),1,1,1,C00135,7 (7),0.0551,0.361,0.215,0.387 Hartnup Disorder,SMPDB,0,,28 (33),1,1,1,C00097,7 (8),0.0551,0.361,0.215,0.387 Renal Glucosuria,SMPDB,0,,28 (33),1,1,1,C00097,7 (8),0.0551,0.361,0.215,0.387 Base-Excision Repair_ AP Site Formation,Reactome,0,,9 (10),1,1,1,C00106,7 (14),0.0551,0.361,0.215,0.387 Kidney Function,SMPDB,0,,28 (33),1,1,1,C00097,7 (8),0.0551,0.361,0.215,0.387 Multiple carboxylase deficiency_ neonatal or early onset form,SMPDB,0,,3 (3),1,1,1,C00047,7 (7),0.0551,0.361,0.215,0.387 glutamine biosynthesis,HumanCyc,0,,1 (1),1,1,1,C00025,7 (7),0.0551,0.361,0.215,0.387 Biotinidase Deficiency,SMPDB,0,,3 (3),1,1,1,C00047,7 (7),0.0551,0.361,0.215,0.387 adenine and adenosine salvage I,HumanCyc,0,,1 (1),1,1,1,C00020,7 (7),0.0551,0.361,0.215,0.387 Propanoate metabolism - Homo sapiens (human),KEGG,1,ACSS2,28 (28),0.0508,1,0,,44 (44),1,1,0.202,0.387 Polycystic Kidney Disease Pathway,Wikipathways,0,,16 (16),1,1,1,C00020,7 (8),0.0551,0.361,0.215,0.387 Arachidonate production from DAG,Reactome,0,,3 (3),1,1,1,C00116,7 (7),0.0551,0.361,0.215,0.387 acetyl-CoA biosynthesis from citrate,HumanCyc,0,,1 (1),1,1,1,C00158,7 (7),0.0551,0.361,0.215,0.387 4-hydroxy-2-nonenal detoxification,HumanCyc,0,,5 (5),1,1,1,C00025,7 (8),0.0551,0.361,0.215,0.387 Triacylglyceride Synthesis,Wikipathways,0,,24 (24),1,1,1,C00116,7 (10),0.0551,0.361,0.215,0.387 glutamate removal from folates,HumanCyc,0,,1 (1),1,1,1,C00025,7 (7),0.0551,0.361,0.215,0.387 AMPK Signaling,Wikipathways,0,,65 (65),1,1,1,C00031,7 (7),0.0551,0.361,0.215,0.387 Effects of Nitric Oxide,Wikipathways,0,,8 (8),1,1,1,C00327,7 (7),0.0551,0.361,0.215,0.387 Polyol Pathway,Wikipathways,0,,4 (4),1,1,1,C00031,7 (7),0.0551,0.361,0.215,0.387 CREB phosphorylation through the activation of Ras,Reactome,0,,28 (32),1,1,1,C00025,7 (7),0.0551,0.361,0.215,0.387 Glucose Transporter Defect (SGLT2),SMPDB,0,,28 (33),1,1,1,C00097,7 (8),0.0551,0.361,0.215,0.387 Signaling by Robo receptor,Reactome,1,EVL,31 (32),0.0561,1,0,,6 (6),1,1,0.218,0.391 Galactose metabolism,EHMN,0,,44 (45),1,1,2,C00031;C00116,48 (48),0.0565,0.368,0.219,0.392 Prostaglandin formation from arachidonate,EHMN,0,,34 (34),1,1,2,C00020;C00116,48 (48),0.0565,0.368,0.219,0.392 GPCR ligand binding,Reactome,0,,412 (454),1,1,3,C00077;C00025;C00047,110 (129),0.0575,0.374,0.222,0.395 Defective ACTH causes Obesity and Pro-opiomelanocortinin deficiency (POMCD),Reactome,0,,412 (454),1,1,3,C00077;C00025;C00047,110 (129),0.0575,0.374,0.222,0.395 Long-term potentiation - Homo sapiens (human),KEGG,0,,65 (67),1,1,1,C00025,7 (7),0.0551,0.361,0.215,0.387 Nicotine addiction - Homo sapiens (human),KEGG,0,,35 (40),1,1,1,C00025,7 (7),0.0551,0.361,0.215,0.387 pyrimidine ribonucleosides degradation,HumanCyc,0,,3 (3),1,1,1,C00106,8 (8),0.0628,0.379,0.237,0.395 Glutamate Binding_ Activation of AMPA Receptors and Synaptic Plasticity,Reactome,0,,28 (31),1,1,1,C00025,8 (8),0.0628,0.379,0.237,0.395 malate-aspartate shuttle,HumanCyc,0,,5 (5),1,1,1,C00025,8 (8),0.0628,0.379,0.237,0.395 Hydroflumethiazide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Conjugation of phenylacetate with glutamine,Reactome,0,,2 (2),1,1,1,C00020,8 (8),0.0628,0.379,0.237,0.395 EPHB-mediated forward signaling,Reactome,0,,42 (47),1,1,1,C00025,8 (8),0.0628,0.379,0.237,0.395 mTOR signalling,Reactome,0,,26 (28),1,1,1,C00020,8 (8),0.0628,0.379,0.237,0.395 EPH-Ephrin signaling,Reactome,0,,77 (83),1,1,1,C00025,8 (8),0.0628,0.379,0.237,0.395 Polythiazide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Methyclothiazide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Bumetanide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Eplerenone Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Triamterene Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Amiloride Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Ethacrynic Acid Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Quinethazone Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Chlorthalidone Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Trichlormethiazide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Furosemide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Homocysteine Degradation,SMPDB,0,,2 (3),1,1,1,C00097,8 (8),0.0628,0.379,0.237,0.395 Gamma-cystathionase deficiency (CTH),SMPDB,0,,2 (3),1,1,1,C00097,8 (8),0.0628,0.379,0.237,0.395 Homocystinuria_ cystathionine beta-synthase deficiency,SMPDB,0,,2 (3),1,1,1,C00097,8 (8),0.0628,0.379,0.237,0.395 Hydrochlorothiazide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Cyclothiazide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Chlorothiazide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Ribosome - Homo sapiens (human),KEGG,1,RPL9,134 (135),0.222,1,0,,0 (0),1,1,0.222,1 SHP2 signaling,PID,0,,57 (61),1,1,1,C00327,8 (8),0.0628,0.379,0.237,0.395 Passive transport by Aquaporins,Reactome,0,,12 (15),1,1,1,C00116,8 (8),0.0628,0.379,0.237,0.395 creatine biosynthesis,HumanCyc,0,,2 (2),1,1,1,C00077,8 (8),0.0628,0.379,0.237,0.395 Endohydrolysis of 1_4-alpha-D-glucosidic linkages in polysaccharides by alpha-amylase,EHMN,0,,6 (6),1,1,1,C01835,8 (8),0.0628,0.379,0.237,0.395 Indapamide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Signaling events mediated by VEGFR1 and VEGFR2,PID,0,,68 (75),1,1,1,C00327,8 (8),0.0628,0.379,0.237,0.395 Organic anion transport,Reactome,0,,5 (5),1,1,1,C00366,8 (8),0.0628,0.379,0.237,0.395 Bendroflumethiazide Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 corticosteroids and cardioprotection,BioCarta,0,,26 (28),1,1,1,C00327,8 (8),0.0628,0.379,0.237,0.395 Metolazone Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Spironolactone Action Pathway,SMPDB,0,,28 (33),1,1,1,C00097,8 (9),0.0628,0.379,0.237,0.395 Synthesis of bile acids and bile salts via 7alpha-hydroxycholesterol,Reactome,0,,24 (24),1,1,2,C00020;C00245,51 (51),0.0629,0.379,0.237,0.395 Folate Metabolism,Wikipathways,0,,65 (66),1,1,2,C00031;C00097,51 (65),0.0629,0.379,0.237,0.395 Fatty acid biosynthesis - Homo sapiens (human),KEGG,0,,13 (13),1,1,2,C01571;C06423,50 (50),0.0608,0.379,0.231,0.395 Base Excision Repair,Wikipathways,0,,0 (0),1,1,1,C00106,9 (13),0.0703,0.401,0.0703,0.401 Generation of second messenger molecules,Reactome,1,EVL,37 (43),0.0666,1,0,,6 (6),1,1,0.247,0.407 Arachidonic acid metabolism,EHMN,0,,96 (101),1,1,2,C00020;C00116,53 (53),0.0673,0.401,0.249,0.407 selenocysteine biosynthesis,HumanCyc,0,,6 (6),1,1,1,C00020,9 (10),0.0703,0.401,0.257,0.407 Glycine_ serine and threonine metabolism - Homo sapiens (human),KEGG,0,,39 (40),1,1,2,C00197;C00097,50 (50),0.0608,0.379,0.231,0.395 Acyl chain remodeling of DAG and TAG,Reactome,0,,5 (5),1,1,1,C00116,9 (9),0.0703,0.401,0.257,0.407 Degradation of GABA,Reactome,0,,2 (2),1,1,1,C00025,9 (9),0.0703,0.401,0.257,0.407 UDP-N-acetyl-D-glucosamine biosynthesis II,HumanCyc,0,,4 (5),1,1,1,C00025,9 (11),0.0703,0.401,0.257,0.407 γ-linolenate biosynthesis,HumanCyc,0,,14 (15),1,1,1,C00020,9 (9),0.0703,0.401,0.257,0.407 PKB-mediated events,Reactome,0,,27 (29),1,1,1,C00020,9 (9),0.0703,0.401,0.257,0.407 7-(3-amino-3-carboxypropyl)-wyosine biosynthesis,HumanCyc,0,,2 (2),1,1,1,C00170,9 (9),0.0703,0.401,0.257,0.407 Lactose Degradation,SMPDB,0,,11 (11),1,1,1,C00031,9 (9),0.0703,0.401,0.257,0.407 Lactose Intolerance,SMPDB,0,,11 (11),1,1,1,C00031,9 (9),0.0703,0.401,0.257,0.407 Thiamine Metabolism,SMPDB,0,,3 (4),1,1,1,C00020,9 (9),0.0703,0.401,0.257,0.407 4-aminobutyrate degradation,HumanCyc,0,,2 (2),1,1,1,C00025,9 (9),0.0703,0.401,0.257,0.407 glutamate biosynthesis/degradation,HumanCyc,0,,3 (3),1,1,1,C00025,9 (9),0.0703,0.401,0.257,0.407 Mismatch repair (MMR) directed by MSH2:MSH6 (MutSalpha),Reactome,0,,15 (15),1,1,1,C00020,9 (9),0.0703,0.401,0.257,0.407 Processive synthesis on the C-strand of the telomere,Reactome,0,,12 (12),1,1,1,C00020,9 (9),0.0703,0.401,0.257,0.407 Mismatch repair (MMR) directed by MSH2:MSH3 (MutSbeta),Reactome,0,,15 (15),1,1,1,C00020,9 (9),0.0703,0.401,0.257,0.407 Mismatch Repair,Reactome,0,,16 (16),1,1,1,C00020,9 (9),0.0703,0.401,0.257,0.407 Cysteine formation from homocysteine,Reactome,0,,2 (3),1,1,1,C00097,9 (9),0.0703,0.401,0.257,0.407 wybutosine biosynthesis,HumanCyc,0,,3 (3),1,1,1,C00170,9 (9),0.0703,0.401,0.257,0.407 Phenylacetate Metabolism,SMPDB,0,,3 (3),1,1,1,C00020,9 (9),0.0703,0.401,0.257,0.407 L-dopa degradation,HumanCyc,0,,1 (1),1,1,1,C00025,9 (12),0.0703,0.401,0.257,0.407 Plasma membrane estrogen receptor signaling,PID,0,,39 (42),1,1,1,C00327,9 (9),0.0703,0.401,0.257,0.407 tRNA splicing,HumanCyc,0,,5 (5),1,1,1,C00020,9 (11),0.0703,0.401,0.257,0.407 Vitamin B1 (thiamin) metabolism,Reactome,0,,3 (4),1,1,1,C00020,9 (9),0.0703,0.401,0.257,0.407 pyrimidine deoxyribonucleosides degradation,HumanCyc,0,,4 (4),1,1,1,C00106,9 (10),0.0703,0.401,0.257,0.407 Post NMDA receptor activation events,Reactome,0,,36 (40),1,1,1,C00025,9 (9),0.0703,0.401,0.257,0.407 phosphatidylethanolamine biosynthesis II,HumanCyc,0,,7 (7),1,1,1,C00346,9 (12),0.0703,0.401,0.257,0.407 Calmodulin induced events,Reactome,0,,26 (28),1,1,1,C00020,9 (9),0.0703,0.401,0.257,0.407 CaM pathway,Reactome,0,,26 (28),1,1,1,C00020,9 (9),0.0703,0.401,0.257,0.407 Sphingolipid metabolism,Wikipathways,0,,0 (0),1,1,2,C00836;C00346,55 (58),0.0719,0.408,0.0719,0.408 Developmental Biology,Reactome,1,EVL,426 (451),0.554,1,1,C00025,17 (18),0.129,0.59,0.26,0.411 Cocaine addiction - Homo sapiens (human),KEGG,0,,48 (50),1,1,1,C00025,8 (8),0.0628,0.379,0.237,0.395 Tyrosine metabolism - Homo sapiens (human),KEGG,1,ADH5,35 (35),0.0631,1,0,,77 (77),1,1,0.237,0.395 Bile acid biosynthesis,EHMN,0,,52 (54),1,1,2,C00020;C00245,55 (55),0.0719,0.408,0.261,0.411 Glycerophospholipid biosynthesis,Wikipathways,0,,0 (0),1,1,2,C00116;C00346,56 (64),0.0741,0.42,0.0741,0.42 Sphingolipid metabolism,Reactome,0,,70 (72),1,1,2,C00836;C00346,56 (58),0.0741,0.42,0.267,0.42 Transport of fatty acids,Reactome,0,,8 (8),1,1,1,C01571,10 (10),0.0779,0.426,0.277,0.424 Base Excision Repair,Reactome,0,,20 (22),1,1,1,C00106,10 (17),0.0779,0.426,0.277,0.424 Glycogen breakdown (glycogenolysis),Reactome,0,,16 (18),1,1,1,C00020,10 (10),0.0779,0.426,0.277,0.424 Long-term depression - Homo sapiens (human),KEGG,0,,59 (60),1,1,1,C00025,9 (9),0.0703,0.401,0.257,0.407 Malate-Aspartate Shuttle,SMPDB,0,,4 (4),1,1,1,C00025,10 (10),0.0779,0.426,0.277,0.424 Pyruvate metabolism - Homo sapiens (human),KEGG,1,ACSS2,40 (40),0.0718,1,0,,31 (31),1,1,0.261,0.411 Putative anti-Inflammatory metabolites formation from EPA,EHMN,0,,41 (42),1,1,1,C00025,10 (10),0.0779,0.426,0.277,0.424 asparagine degradation,HumanCyc,0,,5 (5),1,1,1,C00025,10 (10),0.0779,0.426,0.277,0.424 Interconversion of 2-oxoglutarate and 2-hydroxyglutarate,Reactome,0,,3 (3),1,1,1,C00989,10 (10),0.0779,0.426,0.277,0.424 Pentose and glucuronate interconversions - Homo sapiens (human),KEGG,0,,36 (36),1,1,2,C00379;C00116,55 (55),0.0719,0.408,0.261,0.411 mTOR signaling pathway - Homo sapiens (human),KEGG,0,,59 (60),1,1,1,C00020,10 (10),0.0779,0.426,0.277,0.424 Differentiation Pathway,Wikipathways,0,,44 (44),1,1,1,C00245,10 (13),0.0779,0.426,0.277,0.424 Factors involved in megakaryocyte development and platelet production,Reactome,0,,115 (115),1,1,1,C00020,10 (11),0.0779,0.426,0.277,0.424 Amyotrophic lateral sclerosis (ALS),Wikipathways,0,,33 (35),1,1,1,C00025,10 (10),0.0779,0.426,0.277,0.424 Angiopoietin receptor Tie2-mediated signaling,PID,0,,48 (52),1,1,1,C00327,10 (10),0.0779,0.426,0.277,0.424 uracil degradation,HumanCyc,0,,3 (3),1,1,1,C00106,10 (10),0.0779,0.426,0.277,0.424 Amyotrophic lateral sclerosis (ALS) - Homo sapiens (human),KEGG,0,,47 (51),1,1,1,C00025,10 (10),0.0779,0.426,0.277,0.424 Vitamin B2 (riboflavin) metabolism,Reactome,0,,5 (5),1,1,1,C00020,10 (10),0.0779,0.426,0.277,0.424 HIF1A and PPARG regulation of glycolysis,Wikipathways,0,,0 (0),1,1,1,C00031,10 (10),0.0779,0.426,0.0779,0.426 Signaling by Type 1 Insulin-like Growth Factor 1 Receptor (IGF1R),Wikipathways,0,,0 (0),1,1,1,C00020,10 (10),0.0779,0.426,0.0779,0.426 Synthesis of DNA,Wikipathways,0,,0 (0),1,1,1,C00020,10 (10),0.0779,0.426,0.0779,0.426 Factors involved in megakaryocyte development and platelet production,Wikipathways,0,,0 (0),1,1,1,C00020,10 (11),0.0779,0.426,0.0779,0.426 Alcoholism - Homo sapiens (human),KEGG,0,,175 (180),1,1,1,C00025,10 (10),0.0779,0.426,0.277,0.424 Glycerophospholipid biosynthesis,Reactome,0,,92 (96),1,1,2,C00116;C00346,58 (63),0.0788,0.43,0.279,0.426 Tryptophan Metabolism,SMPDB,0,,19 (19),1,1,2,C00041;C00025,59 (60),0.0812,0.443,0.285,0.434 Amphetamine addiction - Homo sapiens (human),KEGG,0,,65 (68),1,1,1,C00025,10 (10),0.0779,0.426,0.277,0.424 Synthesis of bile acids and bile salts,Reactome,0,,27 (27),1,1,2,C00020;C00245,60 (73),0.0836,0.45,0.291,0.438 UTP and CTP dephosphorylation II,HumanCyc,0,,4 (4),1,1,1,C00025,11 (11),0.0853,0.45,0.295,0.438 PAOs oxidise polyamines to amines,Reactome,0,,2 (2),1,1,1,C00315,11 (11),0.0853,0.45,0.295,0.438 Nitric oxide stimulates guanylate cyclase,Reactome,0,,22 (25),1,1,1,C00327,11 (11),0.0853,0.45,0.295,0.438 Synthesis of UDP-N-acetyl-glucosamine,Reactome,0,,4 (5),1,1,1,C00025,11 (11),0.0853,0.45,0.295,0.438 D-Arginine and D-ornithine metabolism - Homo sapiens (human),KEGG,0,,1 (1),1,1,1,C00077,10 (10),0.0779,0.426,0.277,0.424 superpathway of D-myo-inositol (1_4_5)-trisphosphate metabolism,HumanCyc,0,,19 (20),1,1,1,C03546,11 (11),0.0853,0.45,0.295,0.438 UTP and CTP de novo biosynthesis,HumanCyc,0,,13 (13),1,1,1,C00025,11 (11),0.0853,0.45,0.295,0.438 DARPP-32 events,Reactome,0,,26 (27),1,1,1,C00020,11 (11),0.0853,0.45,0.295,0.438 Fatty acid degradation - Homo sapiens (human),KEGG,1,ADH5,44 (44),0.0787,1,0,,50 (50),1,1,0.279,0.426 Vitamin B1 (thiamin) metabolism,EHMN,0,,3 (4),1,1,1,C00020,11 (11),0.0853,0.45,0.295,0.438 Vitamin H (biotin) metabolism,EHMN,0,,6 (6),1,1,1,C00047,11 (11),0.0853,0.45,0.295,0.438 Herpes simplex infection - Homo sapiens (human),KEGG,1,EEF1D,179 (186),0.285,1,0,,0 (0),1,1,0.285,1 biotin-carboxyl carrier protein assembly,HumanCyc,0,,3 (3),1,1,1,C00020,11 (11),0.0853,0.45,0.295,0.438 Hormone-sensitive lipase (HSL)-mediated triacylglycerol hydrolysis,Reactome,0,,12 (12),1,1,1,C00116,11 (11),0.0853,0.45,0.295,0.438 Activation of NMDA receptor upon glutamate binding and postsynaptic events,Reactome,0,,38 (44),1,1,1,C00025,11 (11),0.0853,0.45,0.295,0.438 β-alanine degradation,HumanCyc,0,,2 (2),1,1,1,C00025,11 (11),0.0853,0.45,0.295,0.438 ion channels and their functional role in vascular endothelium,BioCarta,0,,44 (47),1,1,1,C00327,11 (11),0.0853,0.45,0.295,0.438 Cytosolic sensors of pathogen-associated DNA ,Reactome,1,XRCC6,49 (49),0.0872,1,0,,9 (14),1,1,0.3,0.444 IRS-related events triggered by IGF1R,Reactome,0,,87 (93),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 Signaling by Type 1 Insulin-like Growth Factor 1 Receptor (IGF1R),Reactome,0,,90 (96),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 PLC-gamma1 signalling,Reactome,0,,33 (35),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 EGFR interacts with phospholipase C-gamma,Reactome,0,,33 (35),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 Thromboxane A2 receptor signaling,PID,0,,53 (58),1,1,1,C00327,12 (12),0.0927,0.469,0.313,0.447 D-glucuronate degradation,HumanCyc,0,,2 (2),1,1,1,C00379,12 (13),0.0927,0.469,0.313,0.447 adenine and adenosine salvage III,HumanCyc,0,,3 (3),1,1,1,C00130,12 (12),0.0927,0.469,0.313,0.447 IGF1R signaling cascade,Reactome,0,,90 (96),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 PLCG1 events in ERBB2 signaling,Reactome,0,,34 (36),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 Insulin receptor signalling cascade,Reactome,0,,89 (96),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 PI3K Cascade,Reactome,0,,66 (70),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 IRS-mediated signalling,Reactome,0,,83 (89),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 Valine_ leucine and isoleucine degradation - Homo sapiens (human),KEGG,1,IVD,46 (46),0.0821,1,0,,40 (40),1,1,0.287,0.437 Famotidine Action Pathway,SMPDB,0,,10 (11),1,1,1,C00025,12 (13),0.0927,0.469,0.313,0.447 phytol degradation,HumanCyc,0,,2 (2),1,1,1,C00020,12 (14),0.0927,0.469,0.313,0.447 Ca-dependent events,Reactome,0,,28 (30),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 Lipoate metabolism,EHMN,0,,7 (7),1,1,1,C00047,12 (12),0.0927,0.469,0.313,0.447 Mono-unsaturated fatty acid beta-oxidation,EHMN,0,,21 (22),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 IRS-related events,Reactome,0,,85 (92),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 Gap junction - Homo sapiens (human),KEGG,0,,87 (89),1,1,1,C00025,11 (11),0.0853,0.45,0.295,0.438 DAG and IP3 signaling,Reactome,0,,31 (33),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 Respiratory electron transport,Reactome,0,,86 (92),1,1,1,C00020,12 (14),0.0927,0.469,0.313,0.447 Effects of PIP2 hydrolysis,Reactome,0,,25 (25),1,1,1,C00116,12 (12),0.0927,0.469,0.313,0.447 Sphingolipid Metabolism,Wikipathways,0,,19 (19),1,1,1,C00836,12 (12),0.0927,0.469,0.313,0.447 spermine and spermidine degradation I,HumanCyc,0,,5 (5),1,1,1,C00315,12 (13),0.0927,0.469,0.313,0.447 Endothelin Pathways,Wikipathways,0,,30 (30),1,1,1,C00327,12 (12),0.0927,0.469,0.313,0.447 Transport of vitamins_ nucleosides_ and related molecules,Reactome,0,,39 (39),1,1,2,C01571;C00106,64 (64),0.0933,0.471,0.315,0.449 Effects of PIP2 hydrolysis,Wikipathways,0,,0 (0),1,1,1,C00116,11 (12),0.0853,0.45,0.0853,0.45 Signaling by Insulin receptor,Wikipathways,0,,0 (0),1,1,1,C00020,11 (11),0.0853,0.45,0.0853,0.45 Telomere Maintenance,Wikipathways,0,,0 (0),1,1,1,C00020,11 (11),0.0853,0.45,0.0853,0.45 DAG and IP3 signaling,Wikipathways,0,,0 (0),1,1,1,C00020,11 (11),0.0853,0.45,0.0853,0.45 Bile acid and bile salt metabolism,Reactome,0,,35 (36),1,1,2,C00020;C00245,65 (78),0.0958,0.483,0.321,0.456 Phospholipid metabolism,Reactome,0,,144 (148),1,1,2,C00116;C00346,66 (71),0.0983,0.491,0.326,0.461 tryptophan degradation to 2-amino-3-carboxymuconate semialdehyde,HumanCyc,0,,7 (9),1,1,1,C00041,13 (13),0.1,0.491,0.33,0.461 Circadian entrainment - Homo sapiens (human),KEGG,0,,91 (97),1,1,1,C00025,11 (11),0.0853,0.45,0.295,0.438 CDP-diacylglycerol biosynthesis,HumanCyc,0,,22 (23),1,1,1,C00116,13 (19),0.1,0.491,0.33,0.461 Signaling by Insulin receptor,Reactome,0,,113 (120),1,1,1,C00020,13 (13),0.1,0.491,0.33,0.461 tyrosine degradation,HumanCyc,0,,5 (5),1,1,1,C00025,13 (13),0.1,0.491,0.33,0.461 Lamivudine Metabolism Pathway,SMPDB,0,,18 (18),1,1,1,C00346,13 (13),0.1,0.491,0.33,0.461 GABA shunt,HumanCyc,0,,8 (8),1,1,1,C00025,13 (13),0.1,0.491,0.33,0.461 Methionine salvage pathway,Reactome,0,,6 (6),1,1,1,C00170,13 (17),0.1,0.491,0.33,0.461 Vitamin C metabolism,EHMN,0,,3 (4),1,1,1,C02477,13 (13),0.1,0.491,0.33,0.461 Synthesis of DNA,Reactome,0,,53 (54),1,1,1,C00020,13 (13),0.1,0.491,0.33,0.461 Interconversion of polyamines,Reactome,0,,3 (3),1,1,1,C00315,13 (13),0.1,0.491,0.33,0.461 folate polyglutamylation,HumanCyc,0,,6 (6),1,1,1,C00025,13 (13),0.1,0.491,0.33,0.461 Synthesis of PG,Reactome,0,,9 (9),1,1,1,C00116,13 (13),0.1,0.491,0.33,0.461 Organic anion transporters,Reactome,0,,8 (8),1,1,1,C00025,13 (13),0.1,0.491,0.33,0.461 Transmission across Chemical Synapses,Reactome,1,GLS,195 (213),0.306,1,1,C00025,51 (51),0.34,1,0.34,0.471 Coregulation of Androgen receptor activity,PID,1,XRCC6,60 (63),0.106,1,0,,2 (2),1,1,0.343,0.471 Synthesis of IP2_ IP_ and Ins in the cytosol,Reactome,0,,11 (11),1,1,1,C03546,14 (14),0.107,0.512,0.347,0.471 glutathione-mediated detoxification,HumanCyc,0,,24 (24),1,1,1,C00025,14 (14),0.107,0.512,0.347,0.471 Aflatoxin activation and detoxification,Reactome,0,,22 (25),1,1,1,C00025,14 (25),0.107,0.512,0.347,0.471 Defective ACY1 causes encephalopathy,Reactome,0,,22 (25),1,1,1,C00025,14 (25),0.107,0.512,0.347,0.471 Digestion of dietary carbohydrate,Reactome,0,,10 (10),1,1,1,C01835,14 (14),0.107,0.512,0.347,0.471 eicosapentaenoate biosynthesis,HumanCyc,0,,14 (15),1,1,1,C00020,14 (17),0.107,0.512,0.347,0.471 Telomere C-strand (Lagging Strand) Synthesis,Reactome,0,,23 (23),1,1,1,C00020,14 (14),0.107,0.512,0.347,0.471 Vitamin B2 (riboflavin) metabolism,EHMN,0,,5 (5),1,1,1,C00020,14 (14),0.107,0.512,0.347,0.471 Aquaporin-mediated transport,Reactome,0,,45 (51),1,1,1,C00116,14 (14),0.107,0.512,0.347,0.471 Prolactin signaling pathway - Homo sapiens (human),KEGG,0,,66 (72),1,1,1,C00031,11 (11),0.0853,0.45,0.295,0.438 Extension of Telomeres,Reactome,0,,30 (31),1,1,1,C00020,14 (14),0.107,0.512,0.347,0.471 Telomere Maintenance,Reactome,0,,81 (83),1,1,1,C00020,14 (14),0.107,0.512,0.347,0.471 DNA Replication,Reactome,0,,58 (59),1,1,1,C00020,14 (14),0.107,0.512,0.347,0.471 Glucuronidation,Wikipathways,0,,26 (26),1,1,1,C00031,14 (14),0.107,0.512,0.347,0.471 4-hydroxyproline degradation,HumanCyc,0,,4 (4),1,1,1,C00025,14 (17),0.107,0.512,0.347,0.471 pyrimidine ribonucleosides degradation,HumanCyc,0,,6 (6),1,1,1,C00106,14 (14),0.107,0.512,0.347,0.471 purine ribonucleosides degradation to ribose-1-phosphate,HumanCyc,0,,3 (3),1,1,1,C00385,14 (15),0.107,0.512,0.347,0.471 Interferon Signaling,Reactome,0,,74 (79),1,1,1,C00020,14 (15),0.107,0.512,0.347,0.471 Synaptic vesicle cycle - Homo sapiens (human),KEGG,0,,62 (63),1,1,1,C00025,12 (12),0.0927,0.469,0.313,0.447 Proteasome Degradation,Wikipathways,1,PSMD9,64 (64),0.112,1,0,,1 (1),1,1,0.358,0.485 Glyoxylate metabolism,Reactome,0,,5 (5),1,1,1,C00041,15 (15),0.115,0.54,0.363,0.487 UTP and CTP dephosphorylation I,HumanCyc,0,,3 (3),1,1,1,C00025,15 (15),0.115,0.54,0.363,0.487 Morphine addiction - Homo sapiens (human),KEGG,0,,85 (93),1,1,1,C00020,12 (12),0.0927,0.469,0.313,0.447 Chromosome Maintenance,Reactome,0,,109 (111),1,1,1,C00020,15 (15),0.115,0.54,0.363,0.487 Insulin secretion - Homo sapiens (human),KEGG,0,,85 (86),1,1,1,C00031,13 (13),0.1,0.491,0.33,0.461 Phospholipase C-mediated cascade,Reactome,0,,54 (56),1,1,1,C00020,15 (15),0.115,0.54,0.363,0.487 serotonin degradation,HumanCyc,0,,14 (15),1,1,1,C00097,15 (20),0.115,0.54,0.363,0.487 PTM- gamma carboxylation_ hypusine formation and arylsulfatase activation,Wikipathways,0,,0 (0),1,1,1,C00315,13 (13),0.1,0.491,0.1,0.491 GABA synthesis_ release_ reuptake and degradation,Wikipathways,0,,0 (0),1,1,1,C00025,13 (13),0.1,0.491,0.1,0.491 Cell Cycle_ Mitotic,Reactome,1,SEC13,430 (443),0.558,1,1,C00020,29 (29),0.21,0.847,0.368,0.493 cGMP-PKG signaling pathway - Homo sapiens (human),KEGG,0,,163 (167),1,1,1,C00020,14 (14),0.107,0.512,0.347,0.471 TCR signaling,Reactome,1,EVL,67 (74),0.117,1,0,,11 (11),1,1,0.369,0.493 Regulation of Telomerase,PID,1,XRCC6,68 (71),0.119,1,0,,1 (1),1,1,0.372,0.497 GABA synthesis_ release_ reuptake and degradation,Reactome,0,,20 (21),1,1,1,C00025,16 (16),0.122,0.566,0.378,0.501 VEGFR2 mediated vascular permeability,Reactome,0,,28 (29),1,1,1,C00327,16 (16),0.122,0.566,0.378,0.501 Thyroxine biosynthesis,Reactome,0,,8 (8),1,1,1,C00041,16 (16),0.122,0.566,0.378,0.501 isoleucine degradation,HumanCyc,0,,12 (12),1,1,1,C00025,16 (16),0.122,0.566,0.378,0.501 Omega-3 fatty acid metabolism,EHMN,0,,18 (19),1,1,1,C00020,16 (16),0.122,0.566,0.378,0.501 (S)-reticuline biosynthesis,HumanCyc,0,,1 (1),1,1,1,C00025,16 (16),0.122,0.566,0.378,0.501 S Phase,Wikipathways,0,,0 (0),1,1,1,C00020,14 (14),0.107,0.512,0.107,0.512 Retinol metabolism - Homo sapiens (human),KEGG,1,ADH5,62 (64),0.109,1,0,,24 (24),1,1,0.351,0.476 Phagosomal maturation (early endosomal stage),Reactome,0,,33 (34),1,1,1,C00327,17 (22),0.129,0.59,0.393,0.513 fatty acid α-oxidation,HumanCyc,0,,5 (5),1,1,1,C00020,17 (20),0.129,0.59,0.393,0.513 stearate biosynthesis,HumanCyc,0,,13 (13),1,1,1,C00020,17 (17),0.129,0.59,0.393,0.513 Synthesis of bile acids and bile salts via 24-hydroxycholesterol,Reactome,0,,14 (14),1,1,1,C00020,17 (28),0.129,0.59,0.393,0.513 cysteine biosynthesis,HumanCyc,0,,8 (9),1,1,1,C00097,17 (17),0.129,0.59,0.393,0.513 Synthesis of PE,Reactome,0,,13 (14),1,1,1,C00346,17 (18),0.129,0.59,0.393,0.513 Metabolism of folate and pterines,Reactome,0,,10 (11),1,1,1,C00025,17 (20),0.129,0.59,0.393,0.513 S Phase,Reactome,0,,78 (79),1,1,1,C00020,17 (17),0.129,0.59,0.393,0.513 Digestion of dietary lipid,Reactome,0,,5 (6),1,1,1,C00116,17 (17),0.129,0.59,0.393,0.513 fatty acid α-oxidation III,HumanCyc,0,,3 (3),1,1,1,C00020,17 (23),0.129,0.59,0.393,0.513 Cell Cycle,Reactome,1,SEC13,477 (492),0.596,1,1,C00020,31 (31),0.223,0.892,0.401,0.523 Parkinson_s disease - Homo sapiens (human),KEGG,0,,141 (143),1,1,1,C00020,15 (15),0.115,0.54,0.363,0.487 Neurotransmitter Receptor Binding And Downstream Transmission In The Postsynaptic Cell,Reactome,0,,133 (149),1,1,1,C00025,18 (18),0.136,0.616,0.407,0.527 Transport of nucleosides and free purine and pyrimidine bases across the plasma membrane,Reactome,0,,7 (7),1,1,1,C00106,18 (18),0.136,0.616,0.407,0.527 PLC beta mediated events,Reactome,0,,43 (45),1,1,1,C00020,18 (18),0.136,0.616,0.407,0.527 Amine Oxidase reactions,Reactome,0,,4 (4),1,1,1,C00315,18 (18),0.136,0.616,0.407,0.527 G-protein mediated events,Reactome,0,,44 (46),1,1,1,C00020,18 (18),0.136,0.616,0.407,0.527 Alanine and aspartate metabolism,Wikipathways,0,,12 (12),1,1,1,C00158,18 (18),0.136,0.616,0.407,0.527 Neuronal System,Reactome,1,GLS,274 (294),0.403,1,1,C00025,51 (51),0.34,1,0.409,0.529 HIF-1 signaling pathway - Homo sapiens (human),KEGG,0,,104 (106),1,1,1,C00031,15 (15),0.115,0.54,0.363,0.487 Drug metabolism - cytochrome P450 - Homo sapiens (human),KEGG,1,ADH5,67 (68),0.117,1,0,,88 (88),1,1,0.369,0.493 GLUT-1 deficiency syndrome,SMPDB,0,,11 (11),1,1,1,C00031,19 (19),0.143,0.636,0.421,0.535 Congenital disorder of glycosylation CDG-IId,SMPDB,0,,11 (11),1,1,1,C00031,19 (19),0.143,0.636,0.421,0.535 Lactose Synthesis,SMPDB,0,,11 (11),1,1,1,C00031,19 (19),0.143,0.636,0.421,0.535 Mitochondrial Beta-Oxidation of Short Chain Saturated Fatty Acids,SMPDB,0,,8 (8),1,1,1,C00020,19 (19),0.143,0.636,0.421,0.535 Short-chain 3-hydroxyacyl-CoA dehydrogenase deficiency (SCHAD),SMPDB,0,,8 (8),1,1,1,C00020,19 (19),0.143,0.636,0.421,0.535 Riboflavin Metabolism,SMPDB,0,,5 (5),1,1,1,C00020,19 (19),0.143,0.636,0.421,0.535 Butyrate Metabolism,SMPDB,0,,8 (8),1,1,1,C00020,19 (19),0.143,0.636,0.421,0.535 Di-unsaturated fatty acid beta-oxidation,EHMN,0,,23 (24),1,1,1,C00020,19 (19),0.143,0.636,0.421,0.535 UMP biosynthesis,HumanCyc,0,,2 (2),1,1,1,C00025,19 (20),0.143,0.636,0.421,0.535 Cytokine Signaling in Immune system,Reactome,0,,182 (198),1,1,1,C00020,19 (22),0.143,0.636,0.421,0.535 inosine-5_-phosphate biosynthesis,HumanCyc,0,,3 (4),1,1,1,C00130,19 (19),0.143,0.636,0.421,0.535 Class I MHC mediated antigen processing & presentation,Wikipathways,0,,0 (0),1,1,1,C00020,15 (15),0.115,0.54,0.115,0.54 Metabolism of xenobiotics by cytochrome P450 - Homo sapiens (human),KEGG,1,ADH5,73 (74),0.127,1,0,,121 (121),1,1,0.39,0.513 valine degradation,HumanCyc,0,,13 (13),1,1,1,C00025,20 (20),0.15,0.653,0.434,0.542 Carnitine Synthesis,SMPDB,0,,5 (5),1,1,1,C00047,20 (20),0.15,0.653,0.434,0.542 Respiratory electron transport_ ATP synthesis by chemiosmotic coupling_ and heat production by uncoupling proteins.,Reactome,0,,107 (113),1,1,1,C00020,20 (22),0.15,0.653,0.434,0.542 DAP12 interactions,Reactome,0,,178 (189),1,1,1,C00020,20 (20),0.15,0.653,0.434,0.542 5-aminoimidazole ribonucleotide biosynthesis,HumanCyc,0,,3 (3),1,1,1,C00025,20 (20),0.15,0.653,0.434,0.542 Trans-sulfuration and one carbon metabolism,Wikipathways,0,,31 (31),1,1,1,C00097,20 (20),0.15,0.653,0.434,0.542 Synthesis of very long-chain fatty acyl-CoAs,Reactome,0,,17 (18),1,1,1,C00020,20 (23),0.15,0.653,0.434,0.542 Downstream signaling of activated FGFR,Reactome,0,,145 (155),1,1,1,C00020,20 (20),0.15,0.653,0.434,0.542 Signaling by FGFR,Reactome,0,,158 (168),1,1,1,C00020,20 (20),0.15,0.653,0.434,0.542 DAP12 signaling,Reactome,0,,160 (171),1,1,1,C00020,20 (20),0.15,0.653,0.434,0.542 Downstream signal transduction,Reactome,0,,157 (168),1,1,1,C00020,20 (20),0.15,0.653,0.434,0.542 Alpha-oxidation of phytanate,Reactome,0,,4 (4),1,1,1,C00020,20 (23),0.15,0.653,0.434,0.542 Signaling by ERBB2,Reactome,0,,159 (169),1,1,1,C00020,20 (20),0.15,0.653,0.434,0.542 eNOS activation,Reactome,0,,9 (9),1,1,1,C00327,20 (20),0.15,0.653,0.434,0.542 Chemical carcinogenesis - Homo sapiens (human),KEGG,1,ADH5,78 (80),0.135,1,0,,99 (99),1,1,0.406,0.527 Signaling by FGFR in disease,Reactome,0,,173 (184),1,1,1,C00020,21 (27),0.157,0.677,0.447,0.554 Synthesis of Leukotrienes (LT) and Eoxins (EX),Reactome,0,,24 (27),1,1,1,C00025,21 (28),0.157,0.677,0.447,0.554 Abnormal metabolism in phenylketonuria,Reactome,0,,6 (7),1,1,1,C00041,21 (21),0.157,0.677,0.447,0.554 Signaling by EGFR,Reactome,0,,175 (185),1,1,1,C00020,21 (21),0.157,0.677,0.447,0.554 Pyrimidine salvage reactions,Reactome,0,,10 (10),1,1,1,C00106,21 (21),0.157,0.677,0.447,0.554 tryptophan degradation,HumanCyc,0,,9 (11),1,1,1,C00041,22 (22),0.164,0.694,0.46,0.56 NGF signalling via TRKA from the plasma membrane,Reactome,0,,199 (211),1,1,1,C00020,22 (22),0.164,0.694,0.46,0.56 Familial lipoprotein lipase deficiency,SMPDB,0,,13 (13),1,1,1,C00116,22 (23),0.164,0.694,0.46,0.56 superpathway of pyrimidine ribonucleotides de novo biosynthesis,HumanCyc,0,,15 (15),1,1,1,C00025,22 (23),0.164,0.694,0.46,0.56 Glycerolipid Metabolism,SMPDB,0,,13 (13),1,1,1,C00116,22 (23),0.164,0.694,0.46,0.56 Glycerol Kinase Deficiency,SMPDB,0,,13 (13),1,1,1,C00116,22 (23),0.164,0.694,0.46,0.56 D-glyceric acidura,SMPDB,0,,13 (13),1,1,1,C00116,22 (23),0.164,0.694,0.46,0.56 Phytanic acid peroxisomal oxidation,EHMN,0,,16 (17),1,1,1,C00020,22 (22),0.164,0.694,0.46,0.56 TCA cycle,HumanCyc,0,,19 (22),1,1,1,C00158,22 (23),0.164,0.694,0.46,0.56 Linoleic acid (LA) metabolism,Reactome,0,,7 (7),1,1,1,C00020,22 (24),0.164,0.694,0.46,0.56 Opioid Signalling,Reactome,0,,78 (83),1,1,1,C00020,22 (22),0.164,0.694,0.46,0.56 fatty acid activation,HumanCyc,0,,8 (9),1,1,1,C00020,22 (38),0.164,0.694,0.46,0.56 Signaling by Overexpressed Wild-Type EGFR in Cancer,Reactome,0,,175 (185),1,1,1,C00020,22 (25),0.164,0.694,0.46,0.56 Inositol Phosphate Metabolism,SMPDB,0,,13 (14),1,1,1,C03546,22 (23),0.164,0.694,0.46,0.56 TCA Cycle and PDHc,Wikipathways,0,,0 (0),1,1,1,C00158,16 (16),0.122,0.566,0.122,0.566 Fc epsilon receptor (FCERI) signaling,Wikipathways,0,,0 (0),1,1,1,C00020,16 (16),0.122,0.566,0.122,0.566 Neurotransmitter Receptor Binding And Downstream Transmission In The Postsynaptic Cell,Wikipathways,0,,0 (0),1,1,1,C00025,16 (16),0.122,0.566,0.122,0.566 TCA Cycle,Wikipathways,0,,17 (17),1,1,1,C00158,23 (24),0.17,0.722,0.472,0.574 Omega-6 fatty acid metabolism,EHMN,0,,30 (31),1,1,1,C00020,23 (23),0.17,0.722,0.472,0.574 Resolution of Sister Chromatid Cohesion,Reactome,1,SEC13,101 (102),0.172,1,0,,6 (6),1,1,0.474,0.576 Platelet homeostasis,Reactome,0,,73 (79),1,1,1,C00327,24 (24),0.177,0.742,0.484,0.582 Glycolysis,SMPDB,0,,15 (15),1,1,1,C00031,24 (24),0.177,0.742,0.484,0.582 Glycogenosis_ Type VII. Tarui disease,SMPDB,0,,15 (15),1,1,1,C00031,24 (24),0.177,0.742,0.484,0.582 Mercaptopurine Metabolism Pathway,SMPDB,0,,15 (15),1,1,1,C00020,24 (24),0.177,0.742,0.484,0.582 Linoleate metabolism,EHMN,0,,70 (76),1,1,1,C01601,24 (24),0.177,0.742,0.484,0.582 Signaling by EGFRvIII in Cancer,Reactome,0,,177 (187),1,1,1,C00020,24 (26),0.177,0.742,0.484,0.582 Fanconi-bickel syndrome,SMPDB,0,,15 (15),1,1,1,C00031,24 (24),0.177,0.742,0.484,0.582 Signaling by PDGF,Reactome,0,,178 (191),1,1,1,C00020,24 (24),0.177,0.742,0.484,0.582 Signalling by NGF,Reactome,0,,279 (297),1,1,1,C00020,25 (25),0.184,0.767,0.495,0.594 Mitochondrial Beta-Oxidation of Medium Chain Saturated Fatty Acids,SMPDB,0,,8 (8),1,1,1,C00020,25 (25),0.184,0.767,0.495,0.594 Mitochondrial Beta-Oxidation of Long Chain Saturated Fatty Acids,SMPDB,0,,10 (10),1,1,1,C00020,25 (26),0.184,0.767,0.495,0.594 Mitotic Prometaphase,Reactome,1,SEC13,110 (112),0.186,1,0,,6 (6),1,1,0.498,0.597 VEGFA-VEGFR2 Pathway,Reactome,0,,104 (108),1,1,1,C00327,26 (26),0.19,0.792,0.506,0.605 Signaling by VEGF,Reactome,0,,112 (117),1,1,1,C00327,26 (26),0.19,0.792,0.506,0.605 Bile secretion - Homo sapiens (human),KEGG,0,,72 (72),1,1,3,C00031;C00366;C00315,162 (162),0.14,0.635,0.416,0.535 Nitrogen metabolism - Homo sapiens (human),KEGG,0,,17 (17),1,1,1,C00025,19 (19),0.143,0.636,0.421,0.535 Leukotriene metabolism,EHMN,0,,100 (105),1,1,1,C00020,27 (27),0.197,0.813,0.517,0.613 G alpha (s) signalling events,Reactome,0,,117 (129),1,1,1,C00020,27 (27),0.197,0.813,0.517,0.613 Citric acid cycle (TCA cycle),Reactome,0,,19 (22),1,1,1,C00158,27 (27),0.197,0.813,0.517,0.613 Signaling by EGFR in Cancer,Reactome,0,,177 (187),1,1,1,C00020,27 (34),0.197,0.813,0.517,0.613 Signaling by Ligand-Responsive EGFR Variants in Cancer,Reactome,0,,177 (187),1,1,1,C00020,27 (34),0.197,0.813,0.517,0.613 Respiratory electron transport_ ATP synthesis by chemiosmotic coupling_ and heat production by uncoupling proteins.,Wikipathways,0,,0 (0),1,1,1,C00020,18 (20),0.136,0.616,0.136,0.616 Separation of Sister Chromatids,Reactome,1,SEC13,119 (126),0.199,1,0,,2 (2),1,1,0.521,0.617 MicroRNAs in cancer - Homo sapiens (human),KEGG,1,GLS,291 (297),0.422,1,0,,0 (0),1,1,0.422,1 Folate malabsorption_ hereditary,SMPDB,0,,14 (14),1,1,1,C00025,28 (28),0.203,0.83,0.527,0.618 Citrate cycle (TCA cycle) - Homo sapiens (human),KEGG,0,,30 (30),1,1,1,C00158,20 (20),0.15,0.653,0.434,0.542 Pentose Phosphate Pathway,SMPDB,0,,14 (14),1,1,1,C00020,28 (28),0.203,0.83,0.527,0.618 Homocystinuria due to defect of N(5_10)-methylene THF deficiency,SMPDB,0,,14 (14),1,1,1,C00025,28 (28),0.203,0.83,0.527,0.618 Folate Metabolism,SMPDB,0,,14 (14),1,1,1,C00025,28 (28),0.203,0.83,0.527,0.618 Glucose-6-phosphate dehydrogenase deficiency,SMPDB,0,,14 (14),1,1,1,C00020,28 (28),0.203,0.83,0.527,0.618 Ribose-5-phosphate isomerase deficiency,SMPDB,0,,14 (14),1,1,1,C00020,28 (28),0.203,0.83,0.527,0.618 Transaldolase deficiency,SMPDB,0,,14 (14),1,1,1,C00020,28 (28),0.203,0.83,0.527,0.618 Inositol phosphate metabolism,Reactome,0,,46 (47),1,1,1,C03546,29 (37),0.21,0.847,0.538,0.627 Latent infection of Homo sapiens with Mycobacterium tuberculosis,Reactome,0,,33 (34),1,1,1,C00327,29 (39),0.21,0.847,0.538,0.627 Methotrexate Action Pathway,SMPDB,0,,14 (14),1,1,1,C00025,29 (29),0.21,0.847,0.538,0.627 alpha-linolenic acid (ALA) metabolism,Reactome,0,,12 (12),1,1,1,C00020,29 (32),0.21,0.847,0.538,0.627 Pyrimidine biosynthesis,Reactome,0,,6 (6),1,1,1,C00025,29 (29),0.21,0.847,0.538,0.627 Mitotic Anaphase,Reactome,1,SEC13,130 (137),0.216,1,0,,4 (4),1,1,0.547,0.635 Eukaryotic Translation Termination,Reactome,1,RPL9,130 (139),0.216,1,0,,5 (6),1,1,0.547,0.635 Caffeine metabolism - Homo sapiens (human),KEGG,0,,5 (5),1,1,1,C00385,21 (21),0.157,0.677,0.447,0.554 Inositol Metabolism,SMPDB,0,,21 (22),1,1,1,C03546,30 (30),0.216,0.87,0.548,0.635 Integration of energy metabolism,Reactome,0,,86 (89),1,1,1,C00020,30 (31),0.216,0.87,0.548,0.635 Mitotic Metaphase and Anaphase,Reactome,1,SEC13,131 (138),0.217,1,0,,4 (4),1,1,0.549,0.635 Peptide chain elongation,Reactome,1,RPL9,131 (141),0.217,1,0,,24 (25),1,1,0.549,0.635 Phosphatidylinositol signaling system - Homo sapiens (human),KEGG,0,,81 (81),1,1,1,C03546,27 (27),0.197,0.813,0.517,0.613 Lipid digestion_ mobilization_ and transport,Reactome,0,,50 (53),1,1,1,C00116,31 (32),0.223,0.892,0.557,0.641 Nonsense Mediated Decay (NMD) independent of the Exon Junction Complex (EJC),Reactome,1,RPL9,135 (144),0.223,1,0,,2 (2),1,1,0.558,0.641 Innate Immune System,Reactome,1,XRCC6,579 (607),0.669,1,1,C00020,50 (65),0.335,1,0.559,0.641 Sjogren Larsson Syndrome,SMPDB,0,,16 (17),1,1,1,C00025,32 (32),0.229,0.901,0.567,0.641 Sialuria or French Type Sialuria,SMPDB,0,,16 (17),1,1,1,C00025,32 (32),0.229,0.901,0.567,0.641 Amino Sugar Metabolism,SMPDB,0,,16 (17),1,1,1,C00025,32 (32),0.229,0.901,0.567,0.641 Pyruvate dehydrogenase deficiency (E2),SMPDB,0,,21 (21),1,1,1,C00158,32 (32),0.229,0.901,0.567,0.641 2-ketoglutarate dehydrogenase complex deficiency,SMPDB,0,,21 (21),1,1,1,C00158,32 (32),0.229,0.901,0.567,0.641 Mitochondrial complex II deficiency,SMPDB,0,,21 (21),1,1,1,C00158,32 (32),0.229,0.901,0.567,0.641 Fumarase deficiency,SMPDB,0,,21 (21),1,1,1,C00158,32 (32),0.229,0.901,0.567,0.641 Congenital lactic acidosis,SMPDB,0,,21 (21),1,1,1,C00158,32 (32),0.229,0.901,0.567,0.641 Citric Acid Cycle,SMPDB,0,,21 (21),1,1,1,C00158,32 (32),0.229,0.901,0.567,0.641 Tay-Sachs Disease,SMPDB,0,,16 (17),1,1,1,C00025,32 (32),0.229,0.901,0.567,0.641 Salla Disease/Infantile Sialic Acid Storage Disease,SMPDB,0,,16 (17),1,1,1,C00025,32 (32),0.229,0.901,0.567,0.641 Pyruvate dehydrogenase deficiency (E3),SMPDB,0,,21 (21),1,1,1,C00158,32 (32),0.229,0.901,0.567,0.641 Metabolism of nitric oxide,Reactome,0,,19 (20),1,1,1,C00327,32 (35),0.229,0.901,0.567,0.641 G(M2)-Gangliosidosis: Variant B_ Tay-sachs disease,SMPDB,0,,16 (17),1,1,1,C00025,32 (32),0.229,0.901,0.567,0.641 Platelet activation_ signaling and aggregation,Reactome,0,,212 (231),1,1,1,C00116,32 (32),0.229,0.901,0.567,0.641 eNOS activation and regulation,Reactome,0,,19 (20),1,1,1,C00327,32 (35),0.229,0.901,0.567,0.641 HIV Life Cycle,Reactome,1,XRCC6,142 (147),0.233,1,0,,14 (14),1,1,0.573,0.647 superpathway of pyrimidine deoxyribonucleotides de novo biosynthesis,HumanCyc,0,,22 (22),1,1,1,C00025,33 (34),0.235,0.924,0.576,0.65 Glycogenosis_ Type IA. Von gierke disease,SMPDB,0,,22 (22),1,1,1,C00031,34 (34),0.242,0.938,0.585,0.653 Glycogenosis_ Type IC,SMPDB,0,,22 (22),1,1,1,C00031,34 (34),0.242,0.938,0.585,0.653 Triosephosphate isomerase,SMPDB,0,,22 (22),1,1,1,C00031,34 (34),0.242,0.938,0.585,0.653 Fructose-1_6-diphosphatase deficiency,SMPDB,0,,22 (22),1,1,1,C00031,34 (34),0.242,0.938,0.585,0.653 Fructose and mannose metabolism,EHMN,0,,21 (22),1,1,1,C00031,34 (34),0.242,0.938,0.585,0.653 Gluconeogenesis,SMPDB,0,,22 (22),1,1,1,C00031,34 (34),0.242,0.938,0.585,0.653 Glycogen Storage Disease Type 1A (GSD1A) or Von Gierke Disease,SMPDB,0,,22 (22),1,1,1,C00031,34 (34),0.242,0.938,0.585,0.653 Phosphoenolpyruvate carboxykinase deficiency 1 (PEPCK1),SMPDB,0,,22 (22),1,1,1,C00031,34 (34),0.242,0.938,0.585,0.653 Glycogenosis_ Type IB,SMPDB,0,,22 (22),1,1,1,C00031,34 (34),0.242,0.938,0.585,0.653 Nonsense Mediated Decay (NMD) enhanced by the Exon Junction Complex (EJC),Reactome,1,RPL9,148 (159),0.242,1,0,,4 (4),1,1,0.585,0.653 Nonsense-Mediated Decay (NMD),Reactome,1,RPL9,148 (159),0.242,1,0,,4 (4),1,1,0.585,0.653 Abnormal metabolism in phenylketonuria,Wikipathways,0,,0 (0),1,1,1,C00041,20 (21),0.15,0.653,0.15,0.653 Cytosolic sulfonation of small molecules,Reactome,0,,19 (21),1,1,1,C00020,35 (44),0.248,0.959,0.593,0.66 Selenocompound metabolism - Homo sapiens (human),KEGG,0,,17 (17),1,1,1,C00041,27 (27),0.197,0.813,0.517,0.613 Vitamin B9 (folate) metabolism,EHMN,0,,28 (29),1,1,1,C00025,35 (35),0.248,0.959,0.593,0.66 Processing of Capped Intron-Containing Pre-mRNA,Reactome,1,HNRNPF,155 (162),0.252,1,0,,1 (2),1,1,0.599,0.665 TCA cycle,EHMN,0,,30 (34),1,1,1,C00158,36 (36),0.254,0.98,0.602,0.668 L13a-mediated translational silencing of Ceruloplasmin expression,Reactome,1,RPL9,160 (169),0.259,1,0,,1 (1),1,1,0.609,0.674 3_ -UTR-mediated translational regulation,Reactome,1,RPL9,160 (169),0.259,1,0,,1 (1),1,1,0.609,0.674 Amine-derived hormones,Reactome,0,,16 (16),1,1,1,C00041,37 (38),0.26,1,0.61,0.674 alpha-linolenic (omega3) and linoleic (omega6) acid metabolism,Reactome,0,,12 (12),1,1,1,C00020,37 (42),0.26,1,0.61,0.674 GTP hydrolysis and joining of the 60S ribosomal subunit,Reactome,1,RPL9,161 (170),0.26,1,0,,3 (4),1,1,0.611,0.674 Lysosome - Homo sapiens (human),KEGG,1,NAPSA,120 (122),0.201,1,0,,4 (4),1,1,0.523,0.618 Neurotransmitter Release Cycle,Wikipathways,0,,0 (0),1,1,1,C00025,21 (21),0.157,0.677,0.157,0.677 Opioid Signalling,Wikipathways,0,,0 (0),1,1,1,C00020,21 (21),0.157,0.677,0.157,0.677 Biotin metabolism - Homo sapiens (human),KEGG,0,,3 (3),1,1,1,C00047,28 (28),0.203,0.83,0.527,0.618 Cap-dependent Translation Initiation,Reactome,1,RPL9,167 (177),0.269,1,0,,5 (6),1,1,0.622,0.683 Eukaryotic Translation Initiation,Reactome,1,RPL9,167 (177),0.269,1,0,,5 (6),1,1,0.622,0.683 Thiamine metabolism - Homo sapiens (human),KEGG,0,,3 (4),1,1,1,C00097,30 (30),0.216,0.87,0.548,0.635 Oxidative phosphorylation - Homo sapiens (human),KEGG,1,LHPP,132 (133),0.219,1,0,,16 (16),1,1,0.551,0.637 Peroxisomal lipid metabolism,Reactome,0,,21 (21),1,1,1,C00020,42 (55),0.29,1,0.648,0.71 HIV Infection,Reactome,1,XRCC6,183 (190),0.29,1,0,,16 (16),1,1,0.649,0.71 Long chain acyl-CoA dehydrogenase deficiency (LCAD),SMPDB,0,,14 (14),1,1,1,C00020,43 (43),0.295,1,0.655,0.71 Carnitine palmitoyl transferase deficiency (II),SMPDB,0,,14 (14),1,1,1,C00020,43 (43),0.295,1,0.655,0.71 Very-long-chain acyl coa dehydrogenase deficiency (VLCAD),SMPDB,0,,14 (14),1,1,1,C00020,43 (43),0.295,1,0.655,0.71 Medium chain acyl-coa dehydrogenase deficiency (MCAD),SMPDB,0,,14 (14),1,1,1,C00020,43 (43),0.295,1,0.655,0.71 Short Chain Acyl CoA Dehydrogenase Deficiency (SCAD Deficiency),SMPDB,0,,14 (14),1,1,1,C00020,43 (43),0.295,1,0.655,0.71 Fatty acid Metabolism,SMPDB,0,,14 (14),1,1,1,C00020,43 (43),0.295,1,0.655,0.71 Glutaric Aciduria Type I,SMPDB,0,,14 (14),1,1,1,C00020,43 (43),0.295,1,0.655,0.71 Carnitine palmitoyl transferase deficiency (I),SMPDB,0,,14 (14),1,1,1,C00020,43 (43),0.295,1,0.655,0.71 Trifunctional protein deficiency,SMPDB,0,,14 (14),1,1,1,C00020,43 (43),0.295,1,0.655,0.71 Ethylmalonic Encephalopathy,SMPDB,0,,14 (14),1,1,1,C00020,43 (43),0.295,1,0.655,0.71 Lysine biosynthesis - Homo sapiens (human),KEGG,0,,2 (2),1,1,1,C00047,35 (35),0.248,0.959,0.593,0.66 RNA transport - Homo sapiens (human),KEGG,1,SEC13,162 (165),0.262,1,0,,3 (3),1,1,0.612,0.675 Vitamin digestion and absorption - Homo sapiens (human),KEGG,0,,24 (24),1,1,1,C02477,38 (38),0.266,1,0.618,0.681 Synthesis of substrates in N-glycan biosythesis,Reactome,0,,52 (54),1,1,1,C00025,47 (55),0.318,1,0.682,0.736 Platelet homeostasis,Wikipathways,0,,0 (0),1,1,1,C00327,24 (24),0.177,0.742,0.177,0.742 Neuroactive ligand-receptor interaction - Homo sapiens (human),KEGG,0,,240 (275),1,1,2,C00245;C00025,128 (128),0.276,1,0.632,0.693 N-Glycan biosynthesis,EHMN,0,,43 (43),1,1,1,C00031,51 (51),0.34,1,0.707,0.76 Organelle biogenesis and maintenance,Reactome,0,,293 (307),1,1,1,C00020,52 (53),0.345,1,0.713,0.764 cAMP signaling pathway - Homo sapiens (human),KEGG,0,,189 (200),1,1,1,C00020,40 (40),0.278,1,0.634,0.695 Inositol phosphate metabolism - Homo sapiens (human),KEGG,0,,60 (61),1,1,1,C03546,44 (44),0.301,1,0.662,0.716 Pyrimidine Metabolism,SMPDB,0,,23 (23),1,1,1,C00106,57 (57),0.372,1,0.74,0.789 UMP Synthase Deiciency (Orotic Aciduria),SMPDB,0,,23 (23),1,1,1,C00106,57 (57),0.372,1,0.74,0.789 MNGIE (Mitochondrial Neurogastrointestinal Encephalopathy),SMPDB,0,,23 (23),1,1,1,C00106,57 (57),0.372,1,0.74,0.789 Beta Ureidopropionase Deficiency,SMPDB,0,,23 (23),1,1,1,C00106,57 (57),0.372,1,0.74,0.789 Dihydropyrimidinase Deficiency,SMPDB,0,,23 (23),1,1,1,C00106,57 (57),0.372,1,0.74,0.789 De novo fatty acid biosynthesis,EHMN,0,,44 (44),1,1,1,C00020,59 (59),0.382,1,0.75,0.799 M Phase,Reactome,1,SEC13,260 (269),0.387,1,0,,18 (18),1,1,0.754,0.803 27-Hydroxylase Deficiency,SMPDB,0,,17 (17),1,1,1,C00245,61 (61),0.392,1,0.759,0.803 Congenital Bile Acid Synthesis Defect Type II,SMPDB,0,,17 (17),1,1,1,C00245,61 (61),0.392,1,0.759,0.803 Familial Hypercholanemia (FHCA),SMPDB,0,,17 (17),1,1,1,C00245,61 (61),0.392,1,0.759,0.803 Congenital Bile Acid Synthesis Defect Type III,SMPDB,0,,17 (17),1,1,1,C00245,61 (61),0.392,1,0.759,0.803 Zellweger Syndrome,SMPDB,0,,17 (17),1,1,1,C00245,61 (61),0.392,1,0.759,0.803 Bile Acid Biosynthesis,SMPDB,0,,17 (17),1,1,1,C00245,61 (61),0.392,1,0.759,0.803 Cerebrotendinous Xanthomatosis (CTX),SMPDB,0,,17 (17),1,1,1,C00245,61 (61),0.392,1,0.759,0.803 Biosynthesis of the N-glycan precursor (dolichol lipid-linked oligosaccharide_ LLO) and transfer to a nascent protein,Reactome,0,,68 (70),1,1,1,C00025,62 (70),0.397,1,0.764,0.803 Tyrosinemia Type I,SMPDB,0,,18 (18),1,1,1,C00025,65 (67),0.412,1,0.777,0.803 Tyrosinemia_ transient_ of the newborn,SMPDB,0,,18 (18),1,1,1,C00025,65 (67),0.412,1,0.777,0.803 Dopamine beta-hydroxylase deficiency,SMPDB,0,,18 (18),1,1,1,C00025,65 (67),0.412,1,0.777,0.803 Tyrosine Metabolism,SMPDB,0,,18 (18),1,1,1,C00025,65 (67),0.412,1,0.777,0.803 Alkaptonuria,SMPDB,0,,18 (18),1,1,1,C00025,65 (67),0.412,1,0.777,0.803 Monoamine oxidase-a deficiency (MAO-A),SMPDB,0,,18 (18),1,1,1,C00025,65 (67),0.412,1,0.777,0.803 Hawkinsinuria,SMPDB,0,,18 (18),1,1,1,C00025,65 (67),0.412,1,0.777,0.803 Cyanoamino acid metabolism - Homo sapiens (human),KEGG,0,,7 (7),1,1,1,C00097,45 (45),0.307,1,0.669,0.723 G alpha (i) signalling events,Reactome,0,,220 (242),1,1,1,C00025,67 (83),0.421,1,0.785,0.803 Leukotriene C4 Synthesis Deficiency,SMPDB,0,,30 (30),1,1,1,C00025,69 (69),0.431,1,0.794,0.803 Antrafenine Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,69 (69),0.431,1,0.794,0.803 Lumiracoxib Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,69 (70),0.431,1,0.794,0.803 Nepafenac Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,69 (70),0.431,1,0.794,0.803 Tenoxicam Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,69 (69),0.431,1,0.794,0.803 Arachidonic Acid Metabolism,SMPDB,0,,30 (30),1,1,1,C00025,69 (69),0.431,1,0.794,0.803 Rofecoxib Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,69 (69),0.431,1,0.794,0.803 Trisalicylate-choline Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,69 (69),0.431,1,0.794,0.803 Lornoxicam Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,69 (69),0.431,1,0.794,0.803 Salsalate Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,69 (69),0.431,1,0.794,0.803 Nabumetone Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,69 (70),0.431,1,0.794,0.803 Ketoprofen Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Acetylsalicylic Acid Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Diflunisal Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Acetaminophen Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Sulindac Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Diclofenac Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Ketorolac Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Naproxen Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Etoricoxib Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Carprofen Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Flurbiprofen Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Antipyrine Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Magnesium salicylate Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Phenylbutazone Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Tiaprofenic Acid Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Tolmetin Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Salicylic Acid Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Salicylate-sodium Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Oxaprozin Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Valdecoxib Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Meloxicam Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Suprofen Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Bromfenac Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Mefenamic Acid Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Piroxicam Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Etodolac Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Fenoprofen Action Pathway,SMPDB,0,,30 (30),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Indomethacin Action Pathway,SMPDB,0,,31 (31),1,1,1,C00025,70 (70),0.435,1,0.797,0.803 Primary bile acid biosynthesis - Homo sapiens (human),KEGG,0,,17 (17),1,1,1,C00245,47 (47),0.318,1,0.682,0.736 Celecoxib Action Pathway,SMPDB,0,,36 (36),1,1,1,C00025,73 (73),0.449,1,0.809,0.813 Ibuprofen Action Pathway,SMPDB,0,,49 (50),1,1,1,C00025,74 (74),0.454,1,0.812,0.815 Latent infection of Homo sapiens with Mycobacterium tuberculosis,Wikipathways,0,,0 (0),1,1,1,C00327,28 (40),0.203,0.83,0.203,0.83 Lipid digestion_ mobilization_ and transport,Wikipathways,0,,0 (0),1,1,1,C00116,28 (29),0.203,0.83,0.203,0.83 Arachidonic acid metabolism,Reactome,0,,54 (57),1,1,1,C00025,80 (102),0.48,1,0.832,0.835 Inositol phosphate metabolism,Wikipathways,0,,0 (0),1,1,1,C03546,29 (35),0.21,0.847,0.21,0.847 Integration of energy metabolism,Wikipathways,0,,0 (0),1,1,1,C00020,29 (31),0.21,0.847,0.21,0.847 Starch and sucrose metabolism - Homo sapiens (human),KEGG,0,,56 (56),1,1,1,C00031,51 (51),0.34,1,0.707,0.76 Metabolism of nitric oxide,Wikipathways,0,,0 (0),1,1,1,C00327,31 (35),0.223,0.892,0.223,0.892 Glycerophospholipid metabolism - Homo sapiens (human),KEGG,0,,92 (93),1,1,1,C00346,52 (52),0.345,1,0.713,0.764 Lysine degradation - Homo sapiens (human),KEGG,0,,51 (51),1,1,1,C00047,52 (52),0.345,1,0.713,0.764 Phase 1 - Functionalization of compounds,Wikipathways,0,,0 (0),1,1,2,C00020;C00315,115 (137),0.237,0.93,0.237,0.93 alpha-linolenic (omega3) and linoleic (omega6) acid metabolism,Wikipathways,0,,0 (0),1,1,1,C00020,36 (41),0.254,0.98,0.254,0.98 Signaling events mediated by HDAC Class III,PID,2,XRCC6;ACSS2,39 (39),0.00233,1,0,,0 (0),1,1,0.00233,1 Processing of DNA ends prior to end rejoining,Reactome,1,XRCC6,4 (4),0.00741,1,0,,0 (0),1,1,0.00741,1 Non-homologous end joining,Wikipathways,1,XRCC6,6 (6),0.0111,1,0,,0 (0),1,1,0.0111,1 2-LTR circle formation,Reactome,1,XRCC6,7 (7),0.0129,1,0,,0 (0),1,1,0.0129,1 Integration of provirus,Reactome,1,XRCC6,10 (11),0.0184,1,0,,0 (0),1,1,0.0184,1 Pyrimidine metabolism - Homo sapiens (human),KEGG,0,,103 (105),1,1,1,C00106,66 (66),0.417,1,0.781,0.803 Ifosfamide Pathway_ Pharmacodynamics,PharmGKB,1,ADH5,14 (14),0.0257,1,0,,0 (0),1,1,0.0257,1 Cyclophosphamide Pathway_ Pharmacodynamics,PharmGKB,1,ADH5,17 (17),0.0311,1,0,,0 (0),1,1,0.0311,1 BARD1 signaling events,PID,1,XRCC6,29 (29),0.0525,1,0,,0 (0),1,1,0.0525,1 Translation Factors,Wikipathways,1,EEF1D,49 (50),0.0872,1,0,,0 (0),1,1,0.0872,1 Parkin-Ubiquitin Proteasomal System pathway,Wikipathways,1,PSMD9,72 (73),0.126,1,0,,0 (0),1,1,0.126,1 Phenylalanine metabolism - Homo sapiens (human),KEGG,0,,18 (18),1,1,1,C00180,72 (72),0.445,1,0.805,0.81 Cytoplasmic Ribosomal Proteins,Wikipathways,1,RPL9,88 (88),0.151,1,0,,0 (0),1,1,0.151,1 mRNA Splicing - Major Pathway,Reactome,1,HNRNPF,124 (131),0.207,1,0,,0 (0),1,1,0.207,1 mRNA Splicing,Reactome,1,HNRNPF,124 (131),0.207,1,0,,0 (0),1,1,0.207,1 Ubiquinone and other terpenoid-quinone biosynthesis - Homo sapiens (human),KEGG,0,,10 (10),1,1,1,C02477,90 (90),0.521,1,0.861,0.863 Formation of a pool of free 40S subunits,Reactome,1,RPL9,142 (151),0.233,1,0,,0 (0),1,1,0.233,1 AndrogenReceptor,NetPath,1,XRCC6,145 (149),0.238,1,0,,0 (0),1,1,0.238,1 SRP-dependent cotranslational protein targeting to membrane,Reactome,1,RPL9,154 (164),0.25,1,0,,0 (1),1,1,0.25,1 Amino sugar and nucleotide sugar metabolism - Homo sapiens (human),KEGG,0,,45 (46),1,1,1,C00031,105 (105),0.577,1,0.894,0.895 Peroxisomal lipid metabolism,Wikipathways,0,,0 (0),1,1,1,C00020,42 (55),0.29,1,0.29,1 TCR,NetPath,1,EVL,239 (251),0.362,1,0,,0 (0),1,1,0.362,1 Bile acid and bile salt metabolism,Wikipathways,0,,0 (0),1,1,1,C00020,56 (70),0.367,1,0.367,1 Asparagine N-linked glycosylation,Wikipathways,0,,0 (0),1,1,1,C00025,61 (64),0.392,1,0.392,1 Arachidonic acid metabolism,Wikipathways,0,,0 (0),1,1,1,C00025,63 (78),0.402,1,0.402,1 Porphyrin and chlorophyll metabolism - Homo sapiens (human),KEGG,0,,41 (41),1,1,1,C00025,124 (124),0.639,1,0.925,0.925 miR-targeted genes in muscle cell - TarBase,Wikipathways,1,ABHD10,398 (398),0.529,1,0,,0 (0),1,1,0.529,1 miR-targeted genes in lymphocytes - TarBase,Wikipathways,1,ABHD10,482 (482),0.6,1,0,,0 (0),1,1,0.6,1 ================================================ FILE: Demos/Data Analysis Workflow/results/statistical_results.csv ================================================ "","ID","description","type","healthy.mean.....std.dev","sick.mean.....std.dev","mean.sick_mean.healthy","group_p.values","group_adjusted.p.values","group_q.values" "1","C00379","xylitol","metabolite","579 +/- 300","880 +/- 410",1.52,0.00328545050468262,0.0361722723945709,0.0159161803445209 "2","C00385","xanthine","metabolite","574 +/- 350","1670 +/- 1500",2.91,0.00039090476261604,0.013030158753868,0.00607284567246352 "3","C00105 ","uridine-5'-monophosphate","metabolite","120 +/- 55","131 +/- 87",1.08,0.610200483328374,0.729045221914476,0.357776810754231 "4","C00299","uridine","metabolite","1990 +/- 1100","1550 +/- 1300",0.78,0.18894602157576,0.374150537773782,0.171821768581724 "5","C00366","uric acid","metabolite","2960 +/- 2100","4150 +/- 1900",1.4,0.0366106802463671,0.124104000835143,0.0592287600595077 "6","C00086","urea","metabolite","104000 +/- 42000","98900 +/- 38000",0.95,0.669390202042164,0.773861505251056,0.379317299747421 "7","C00106","uracil","metabolite","16300 +/- 11000","24700 +/- 17000",1.52,0.0334733584315987,0.123975401598514,0.0574215010905176 "8","C00082","tyrosine","metabolite","37200 +/- 16000","42700 +/- 17000",1.15,0.225792163316927,0.40320029163737,0.190812415121345 "9","C00078","tryptophan","metabolite","8010 +/- 5800","7120 +/- 3400",0.89,0.502255888686614,0.643647776650811,0.314383605053178 "10","C01083","trehalose","metabolite","3040 +/- 3000","6020 +/- 7200",1.98,0.0536152999524518,0.167444393629659,0.0757129572671398 "11","C01157","trans-4-hydroxyproline","metabolite","394 +/- 120","467 +/- 170",1.18,0.0758355243406092,0.204856156963675,0.0948887416739869 "12","C02477","tocopherol alpha","metabolite","1160 +/- 1400","2600 +/- 1200",2.24,0.000150503421423003,0.00771455795526333,0.0036324624978498 "13","C00178","thymine","metabolite","148 +/- 42","159 +/- 72",1.08,0.474287088571268,0.628097012342116,0.302166723515868 "14","C00188","threonine","metabolite","32400 +/- 9600","34800 +/- 8700",1.08,0.331102583126561,0.50167058049479,0.239100214453653 "15","C01620","threonic acid","metabolite","6560 +/- 2500","6920 +/- 4500",1.06,0.713870260961445,0.803136066730948,0.394577053665743 "16","C00245","taurine","metabolite","93000 +/- 29000","125000 +/- 59000",1.34,0.0142252181612132,0.0812869609212182,0.0387942844063903 "17","D09007","tagatose","metabolite","163 +/- 61","108 +/- 56",0.66,0.00110565990755587,0.023396520705865,0.0101120604656123 "18","C01530","stearic acid","metabolite","246000 +/- 110000","219000 +/- 150000",0.89,0.466703534304084,0.622271379072112,0.29877880335116 "19","C00315","spermidine","metabolite","276 +/- 100","189 +/- 110",0.68,0.00376483482897061,0.0361722723945709,0.016878366812324 "20","C00794","sorbitol","metabolite","454 +/- 470","535 +/- 730",1.18,0.6284298127158,0.73073234036721,0.364568519605175 "21","C00493","shikimic acid","metabolite","292 +/- 130","306 +/- 180",1.05,0.736754854822454,0.81695201055171,0.402139384789744 "22","C00121","ribose","metabolite","726 +/- 620","698 +/- 440",0.96,0.85143234827253,0.883370899047943,0.4373573640815 "23","C00013","pyrophosphate","metabolite","441 +/- 240","461 +/- 260",1.05,0.766763163297035,0.833205495903235,0.411774108279692 "24","C00134","putrescine","metabolite","384 +/- 280","423 +/- 290",1.1,0.619398259889562,0.729045221914476,0.36122167938282 "25","C02067","pseudo uridine","metabolite","925 +/- 300","886 +/- 280",0.96,0.619688438627305,0.729045221914476,0.361329759571424 "26","C00148","proline","metabolite","68200 +/- 38000","94500 +/- 49000",1.39,0.0311610632409306,0.120715763058434,0.0559433595293094 "27","C00408","pipecolic acid","metabolite","224 +/- 50","235 +/- 74",1.05,0.509346910367233,0.643647776650811,0.317413317953474 "28","C00346","phosphoethanolamine","metabolite","13500 +/- 7700","21200 +/- 14000",1.57,0.013411529593671,0.0810371067971812,0.0376724937670257 "29","C00166","phenylpyruvic acid","metabolite","396 +/- 120","433 +/- 97",1.09,0.223180529519009,0.40320029163737,0.189555230532812 "30","C01601","pelargonic acid","metabolite","18800 +/- 8300","13900 +/- 6700",0.74,0.0210762888444018,0.100363280211437,0.0474777261697044 "31","C00864","pantothenic acid","metabolite","1200 +/- 900","975 +/- 340",0.81,0.222687705090258,0.40320029163737,0.189316573834548 "32","C00249","palmitic acid","metabolite","40000 +/- 13000","36300 +/- 19000",0.91,0.410550185349807,0.570208590763622,0.272630754320155 "33","C01879","oxoproline","metabolite","96200 +/- 18000","110000 +/- 33000",1.15,0.0579236291551109,0.169178754049596,0.0793887501329705 "34","C00295","orotic acid","metabolite","186 +/- 140","211 +/- 210",1.13,0.612593773617059,0.729045221914476,0.358676747543567 "35","C00077","ornithine","metabolite","7300 +/- 2900","3910 +/- 2100",0.54,9.43331694404056e-06,0.00188666338880811,0.000900412012641018 "36","D01924","octadecanol","metabolite","1620 +/- 5100","514 +/- 380",0.32,0.266215262893444,0.451212309988888,0.210985145434877 "37","C06424","myristic acid","metabolite","2310 +/- 520","2080 +/- 880",0.9,0.257023581181974,0.443144105486162,0.206552985715321 "38","C00073","methionine","metabolite","9880 +/- 4900","11800 +/- 6100",1.19,0.216498165456674,0.40320029163737,0.186280079508244 "39","C00159","mannose","metabolite","3420 +/- 2000","3290 +/- 2500",0.96,0.830774157508609,0.871885795193779,0.431322821154691 "40","C01835","maltotriose","metabolite","3290 +/- 2200","7370 +/- 9700",2.24,0.0376626546976765,0.125542182325588,0.059915133632047 "41","C00208","maltose","metabolite","2390 +/- 2200","4590 +/- 5400",1.92,0.0544194279296391,0.167444393629659,0.0764158574242376 "42","C07272","maleimide","metabolite","626 +/- 330","644 +/- 370",1.03,0.856869772076505,0.883370899047943,0.438924478848032 "43","C00047","lysine","metabolite","24900 +/- 25000","10700 +/- 12000",0.43,0.00996149512009512,0.071153536572108,0.0320594353178956 "44","C01595","linoleic acid","metabolite","1210 +/- 750","1310 +/- 600",1.08,0.615589596542029,0.729045221914476,0.3597997033252 "45","C00123","leucine","metabolite","107000 +/- 47000","125000 +/- 66000",1.16,0.265781453888122,0.451212309988888,0.210778720302186 "46","C02679 ","lauric acid","metabolite","8490 +/- 3500","7000 +/- 2700",0.82,0.086433883256494,0.217685164150831,0.10340021807679 "47","C00328","kynurenine","metabolite","826 +/- 620","759 +/- 740",0.92,0.718806779724198,0.803136066730948,0.39622449068396 "48","C03546","inositol-4-monophosphate","metabolite","302 +/- 130","218 +/- 120",0.72,0.0164080411102474,0.0886921141094456,0.041912480929005 "49","C00137","inositol myo-","metabolite","260000 +/- 85000","260000 +/- 120000",1,0.991599588771298,0.991599588771298,0.475146821165374 "50","C00130","inosine 5'-monophosphate","metabolite","199 +/- 340","625 +/- 790",3.14,0.0130565042625477,0.0810371067971812,0.0371616684439078 "51","C00294","inosine","metabolite","36000 +/- 21000","39500 +/- 18000",1.1,0.505192752792119,0.643647776650811,0.315641667685943 "52","C02043","indole-3-lactate","metabolite","243 +/- 98","293 +/- 180",1.21,0.203395727334973,0.391145629490332,0.179603606846582 "53","C16526","icosenoic acid","metabolite","663 +/- 260","632 +/- 340",0.95,0.7120111208497,0.803136066730948,0.393954280279788 "54","C00262","hypoxanthine","metabolite","45000 +/- 18000","54100 +/- 23000",1.2,0.113813847243337,0.258667834643947,0.123278102847589 "55","C00192","hydroxylamine","metabolite","7640 +/- 2900","6050 +/- 4000",0.79,0.100469887508766,0.239214017878015,0.113797249879282 "56","C00135","histidine","metabolite","6550 +/- 3400","4720 +/- 1700",0.72,0.0157996448519427,0.087775804733015,0.0410838025555293 "57","C00160","glycolic acid","metabolite","1360 +/- 390","1250 +/- 580",0.92,0.399296554607556,0.562085005266816,0.267264577851748 "58","C00037","glycine","metabolite","338000 +/- 64000","373000 +/- 1e+05",1.11,0.129084938984622,0.283703162603566,0.134305825956803 "59","C00093","glycerol-alpha-phosphate","metabolite","7840 +/- 10000","9120 +/- 6100",1.16,0.580341330408203,0.707733329766101,0.346332919078033 "60","C05401","glycerol-3-galactoside","metabolite","361 +/- 370","695 +/- 1200",1.93,0.161118623692168,0.332203347818903,0.155447545885591 "61","C00116","glycerol","metabolite","22900 +/- 7600","27700 +/- 8200",1.21,0.0270207871502094,0.116240884373239,0.0529215876871668 "62","C00064","glutamine","metabolite","11600 +/- 3000","11800 +/- 3800",1.02,0.770715083710493,0.833205495903235,0.41301985397975 "63","C00025","glutamic acid","metabolite","72900 +/- 18000","96200 +/- 30000",1.32,0.00111252948326934,0.023396520705865,0.0101348246975382 "64","C00191","glucuronic acid","metabolite","229 +/- 110","207 +/- 110",0.9,0.477353729380008,0.628097012342116,0.303527454411203 "65","C00031","glucose","metabolite","335000 +/- 2e+05","144000 +/- 140000",0.43,0.000154291159105267,0.00771455795526333,0.00365064633329483 "66","C00257","gluconic acid","metabolite","502 +/- 580","199 +/- 120",0.4,0.0105176944711578,0.072535823939019,0.0330724274491998 "67","C00446","galactose-6-phosphate","metabolite","221 +/- 110","197 +/- 160",0.89,0.516061954477097,0.643647776650811,0.320257808963803 "68","C00122","fumaric acid","metabolite","4470 +/- 1900","4250 +/- 2100",0.95,0.686705360754177,0.784806126576202,0.385348155828769 "69","C01018","fucose+ rhamnose","metabolite","1150 +/- 730","2500 +/- 2000",2.17,0.00175499678084269,0.024873584628788,0.0116847020661663 "70","C01094","fructose 1 phosphate","metabolite","168 +/- 74","159 +/- 88",0.94,0.675649390425982,0.776608494742508,0.381510979950102 "71","C00095","fructose","metabolite","685 +/- 800","934 +/- 1100",1.36,0.34938837100958,0.525396046630947,0.246143144304717 "72","C00836","dihydrosphingosine","metabolite","341 +/- 150","234 +/- 210",0.69,0.0360026114833942,0.124104000835143,0.0588947333343749 "73","C00112","cytidine-5'-diphosphate","metabolite","5290 +/- 3600","5310 +/- 3900",1,0.984486499102861,0.989433667440062,0.473351796665577 "74","C00097","cysteine","metabolite","5160 +/- 2500","8460 +/- 3400",1.64,0.00015078430404477,0.00771455795526333,0.0036338359417622 "75","C00791","creatinine","metabolite","10800 +/- 4200","15500 +/- 6600",1.44,0.00288081460529995,0.0347878155376343,0.0149745227503144 "76","C00327","citrulline","metabolite","597 +/- 240","401 +/- 180",0.67,0.00162923691829092,0.024873584628788,0.0114506409842931 "77","C00158","citric acid","metabolite","3640 +/- 2000","1930 +/- 1400",0.53,0.000579756071214233,0.0165644591775495,0.00760320297361104 "78","C00187","cholesterol","metabolite","132000 +/- 47000","126000 +/- 48000",0.95,0.62590607396611,0.73073234036721,0.363636828781049 "79","C06423","caprylic acid","metabolite","1470 +/- 530","1080 +/- 490",0.74,0.00734854914898753,0.0565273011460579,0.026584412159079 "80","C01571","capric acid","metabolite","2800 +/- 830","2240 +/- 1100",0.8,0.0363759041820177,0.124104000835143,0.0591006690983508 "81","C06555","biuret","metabolite","269 +/- 180","150 +/- 66",0.56,0.00185532094508146,0.024873584628788,0.0118544519881803 "82","C00180","benzoic acid","metabolite","12200 +/- 3100","10000 +/- 4400",0.83,0.0468593104881019,0.148759715835244,0.0694773959805803 "83","C08281","behenic acid","metabolite","990 +/- 460","821 +/- 430",0.83,0.168264576786352,0.339928437952225,0.159839930977894 "84","C00152","asparagine","metabolite","6100 +/- 1800","7320 +/- 3300",1.2,0.096381547589217,0.235076945339554,0.110881902747376 "85","C06425","arachidic acid","metabolite","3880 +/- 2400","3350 +/- 2200",0.87,0.401890778765773,0.562085005266816,0.268423912079844 "86","C00026","alpha ketoglutaric acid","metabolite","114 +/- 68","109 +/- 81",0.95,0.782516107944086,0.836915623469611,0.416708630314735 "87","C00041","alanine","metabolite","213000 +/- 59000","258000 +/- 86000",1.21,0.0296848111802146,0.119912537411498,0.0549254772257551 "88","C06104","adipic acid","metabolite","926 +/- 680","591 +/- 370",0.64,0.029795109415341,0.119912537411498,0.0550036789067433 "89","C00020","adenosine-5-phosphate","metabolite","1300 +/- 1500","3020 +/- 2800",2.32,0.00683740192348928,0.0546992153879142,0.0253505504669233 "90","C00212","adenosine","metabolite","2690 +/- 3600","2650 +/- 2100",0.98,0.954861765806219,0.973547556674085,0.465741880978428 "91","C05659","5-methoxytryptamine","metabolite","204 +/- 120","164 +/- 100",0.8,0.195677615445374,0.383681598912498,0.175504023097903 "92","C00170","5'-deoxy-5'-methylthioadenosine","metabolite","216 +/- 83","355 +/- 160",1.64,0.000200405517131727,0.0080162206852691,0.00382574943883209 "93","C00431","5-aminovaleric acid","metabolite","451 +/- 140","497 +/- 200",1.1,0.32482727779475,0.495919508083587,0.236601251638133 "94","C00989","4-hydroxybutyric acid","metabolite","209 +/- 42","282 +/- 110",1.35,0.00295696432069891,0.0347878155376343,0.0151620939094918 "95","C00197","3-phosphoglycerate","metabolite","1030 +/- 660","510 +/- 490",0.49,0.0018655188471591,0.024873584628788,0.0118709434497229 "96","C05145","3-aminoisobutyric acid","metabolite","1770 +/- 540","1600 +/- 830",0.9,0.35445072268183,0.529030929375865,0.248032559609213 "97","C02112","2-monoolein","metabolite","326 +/- 170","341 +/- 190",1.04,0.776081010597322,0.834495710319702,0.414702908509561 "98","C00233","2-ketoisocaproic acid","metabolite","487 +/- 270","358 +/- 240",0.74,0.071356058130456,0.195496049672482,0.0910515660134888 "99","C00956","2-aminoadipic acid","metabolite","200 +/- 130","207 +/- 220",1.03,0.893520660904468,0.916431447081505,0.449264059145774 "100","C07326","1,5-anhydroglucitol","metabolite","7390 +/- 4300","5940 +/- 3600",0.8,0.185357340547488,0.370714681094977,0.169816580435001 "101","SPTAN1","spectrin, alpha, non-erythrocytic 1","protein","5.29e+09 +/- 3.4e+09","3.77e+09 +/- 2.2e+09",0.71,0.0586164101090191,0.169178754049596,0.0799597140126616 "102","CFH","complement factor H","protein","1.85e+08 +/- 2e+08","2.08e+08 +/- 2.6e+08",1.12,0.715899490472791,0.803136066730948,0.395255344181119 "103","VPS13C","vacuolar protein sorting 13 homolog C (S. cerevisiae)","protein","17800000 +/- 1.7e+07","24500000 +/- 2.5e+07",1.37,0.252402218773571,0.438960380475775,0.204277371609485 "104","XRCC6","X-ray repair complementing defective repair in Chinese hamster cells 6","protein","2.51e+08 +/- 3.4e+08","7.33e+08 +/- 9.1e+08",2.92,0.0130794769787721,0.0810371067971812,0.0371951322316488 "105","APOA1","apolipoprotein A-I","protein","4.58e+09 +/- 4.5e+09","4.01e+09 +/- 4e+09",0.87,0.616485245952436,0.729045221914476,0.360134665799367 "106","SUPT16H","suppressor of Ty 16 homolog (S. cerevisiae)","protein","11800000 +/- 3.6e+07","2.4e+07 +/- 4.5e+07",2.04,0.277201453442009,0.458494327452479,0.216124608658543 "107","ACTBL2","actin, beta-like 2","protein","6.72e+09 +/- 5.7e+09","7.99e+09 +/- 6.7e+09",1.19,0.454902619414127,0.613690977390211,0.293440873249496 "108","HPR","haptoglobin-related protein","protein","1.84e+09 +/- 2.2e+09","1.81e+09 +/- 2.3e+09",0.98,0.958944343323973,0.973547556674085,0.466803641085138 "109","APOE","apolipoprotein E","protein","5.29e+08 +/- 4.3e+08","4.25e+08 +/- 3.2e+08",0.8,0.315379656348093,0.495919508083587,0.232756239848219 "110","MAOB","monoamine oxidase B","protein","2.44e+08 +/- 1.5e+08","2.57e+08 +/- 2e+08",1.05,0.7984770882671,0.849443710922447,0.421624593129142 "111","DMD","dystrophin","protein","5030000 +/- 6800000","6690000 +/- 1.1e+07",1.33,0.501339399716578,0.643647776650811,0.313990061771896 "112","EEF1D","eukaryotic translation elongation factor 1 delta (guanine nucleotide exchange protein)","protein","2.12e+08 +/- 2.6e+08","5.18e+08 +/- 5.7e+08",2.45,0.0137763081555208,0.0810371067971812,0.0381836082789274 "113","GLS","glutaminase","protein","13900000 +/- 3.1e+07","83700000 +/- 1.5e+08",6.04,0.0255208160350274,0.113425849044566,0.0516866006445412 "114","NAPSA","napsin A aspartic peptidase","protein","3.33e+08 +/- 3.2e+08","1.37e+09 +/- 2.2e+09",4.11,0.0200583572485543,0.0987772806835311,0.0463772386731902 "115","EIF3L","eukaryotic translation initiation factor 3, subunit L","protein","23500000 +/- 2.7e+07","58400000 +/- 1e+08",2.49,0.0843549281697272,0.216294687614685,0.101790075381712 "116","LTBP4","latent transforming growth factor beta binding protein 4","protein","4540000 +/- 7100000","10200000 +/- 1.9e+07",2.24,0.158695968958725,0.330616601997343,0.153927039640461 "117","CNP","2',3'-cyclic nucleotide 3' phosphodiesterase","protein","35700000 +/- 6.4e+07","49700000 +/- 6.3e+07",1.39,0.423472170536489,0.584099545567571,0.278819205711635 "118","ANXA7","annexin A7","protein","81800000 +/- 7.5e+07","1.54e+08 +/- 1.7e+08",1.88,0.0434724943620194,0.140233852780708,0.0661110685594147 "119","ADH5","alcohol dehydrogenase 5 (class III), chi polypeptide","protein","1.18e+08 +/- 1.1e+08","2.33e+08 +/- 2.2e+08",1.98,0.0184761354568517,0.0947494125992395,0.0445464424868661 "120","GNA11","guanine nucleotide binding protein (G protein), alpha 11 (Gq class)","protein","66400000 +/- 6.7e+07","61500000 +/- 8.3e+07",0.93,0.813525418806749,0.860873459054761,0.426184167722323 "121","DDX21","DEAD (Asp-Glu-Ala-Asp) box helicase 21","protein","6410000 +/- 2e+07","19700000 +/- 5.3e+07",3.07,0.228400677979968,0.404248987575164,0.192055584005662 "122","TAP2","transporter 2, ATP-binding cassette, sub-family B (MDR/TAP)","protein","19500000 +/- 2.5e+07","38200000 +/- 6.5e+07",1.96,0.167519784098847,0.339928437952225,0.159388461599096 "123","SLFN5","schlafen family member 5","protein","10100000 +/- 1.4e+07","2.1e+07 +/- 2.3e+07",2.07,0.0429044717342592,0.140233852780708,0.0655294786176391 "124","HTRA1","HtrA serine peptidase 1","protein","1.29e+08 +/- 1.1e+08","1.35e+08 +/- 1.3e+08",1.05,0.856207828723303,0.883370899047943,0.438734167620614 "125","ANKFY1","ankyrin repeat and FYVE domain containing 1","protein","9590000 +/- 1e+07","11600000 +/- 1.5e+07",1.21,0.566022044749419,0.694505576379655,0.340698937782297 "126","HNRNPF","heterogeneous nuclear ribonucleoprotein F","protein","1.59e+08 +/- 2.4e+08","4.12e+08 +/- 4.2e+08",2.59,0.00853303001599452,0.0632076297481076,0.0292256876657171 "127","PC","pyruvate carboxylase","protein","1170000 +/- 4700000","8050000 +/- 2e+07",6.88,0.0909259330204405,0.22450847659368,0.106845812702126 "128","SEC24C","SEC24 family member C","protein","14500000 +/- 2.2e+07","38500000 +/- 4.8e+07",2.66,0.0223397627775472,0.103905873383941,0.0487673830313929 "129","IVD","isovaleryl-CoA dehydrogenase","protein","1.1e+07 +/- 1.4e+07","58900000 +/- 7.1e+07",5.34,0.00116982603529325,0.023396520705865,0.0103179316536522 "130","ENTPD1","ectonucleoside triphosphate diphosphohydrolase 1","protein","50800000 +/- 5.3e+07","40100000 +/- 6.5e+07",0.79,0.506399141816642,0.643647776650811,0.316157111023143 "131","ACSS2","acyl-CoA synthetase short-chain family member 2","protein","698000 +/- 2e+06","12700000 +/- 2.8e+07",18.25,0.0318359195600991,0.120715763058434,0.0563887944164092 "132","API5","apoptosis inhibitor 5","protein","16700000 +/- 3.2e+07","39900000 +/- 4.2e+07",2.39,0.0254517405839864,0.113425849044566,0.0516276981564353 "133","RPL9","ribosomal protein L9","protein","96200000 +/- 1e+08","1.88e+08 +/- 1.8e+08",1.96,0.0273166078277111,0.116240884373239,0.0531556044202828 "134","GLOD4","glyoxalase domain containing 4","protein","24500000 +/- 4.5e+07","1.24e+08 +/- 1.4e+08",5.06,0.00129696068640331,0.0235811033891511,0.0106853821389526 "135","ABHD10","abhydrolase domain containing 10","protein","2530000 +/- 6700000","20100000 +/- 4.2e+07",7.93,0.0351529101300627,0.124104000835143,0.0584152922366447 "136","F11R","F11 receptor","protein","48500000 +/- 6.5e+07","3.2e+07 +/- 2.9e+07",0.66,0.230645819902185,0.404641789302078,0.193115703662602 "137","STXBP1","syntaxin binding protein 1","protein","4300000 +/- 6900000","3570000 +/- 1e+07",0.83,0.759900165465411,0.831117224108447,0.409598105735211 "138","FER1L6","fer-1-like family member 6","protein","0 +/- 0","47200 +/- 250000",Inf,0.321941348172835,0.495919508083587,0.235437415083026 "139","SCYL1","SCY1-like 1 (S. cerevisiae)","protein","749000 +/- 2700000","3240000 +/- 7200000",4.32,0.0996681227865923,0.239214017878015,0.113232652513425 "140","GBP2","guanylate binding protein 2, interferon-inducible","protein","7090000 +/- 1.2e+07","53900000 +/- 1.6e+08",7.6,0.131333557323075,0.285507733311033,0.13586363143613 "141","PSMD7","proteasome (prosome, macropain) 26S subunit, non-ATPase, 7","protein","67400000 +/- 1.6e+08","96200000 +/- 1.7e+08",1.43,0.521637686952684,0.64399714438603,0.322601734946892 "142","GMDS","GDP-mannose 4,6-dehydratase","protein","89100 +/- 460000","5580000 +/- 1.5e+07",62.63,0.0592125639173586,0.169178754049596,0.0804467515819778 "143","SIAE","sialic acid acetylesterase","protein","67800 +/- 350000","6710000 +/- 1.9e+07",98.86,0.0827911528953708,0.215041955572392,0.100560375510433 "144","TXNL1","thioredoxin-like 1","protein","12400000 +/- 2.2e+07","46100000 +/- 5.4e+07",3.73,0.00380982884819193,0.0361722723945709,0.0169613598162995 "145","EPS8","epidermal growth factor receptor pathway substrate 8","protein","411000 +/- 1300000","6240000 +/- 1.1e+07",15.2,0.0110649381689454,0.0737662544596362,0.0340247204561482 "146","U2SURP","U2 snRNP-associated SURP domain containing","protein","2420000 +/- 4700000","4410000 +/- 8100000",1.83,0.275552916683402,0.458494327452479,0.215364074736382 "147","AGO2","argonaute RISC catalytic component 2","protein","2490000 +/- 6700000","4990000 +/- 7900000",2.01,0.213894769544741,0.40320029163737,0.184980842044265 "148","DTX3L","deltex 3 like, E3 ubiquitin ligase","protein","4520000 +/- 4900000","7510000 +/- 1.7e+07",1.66,0.379738830717982,0.542484043882832,0.258361323092 "149","EVL","Enah/Vasp-like","protein","6110000 +/- 8600000","21900000 +/- 3.7e+07",3.59,0.0349413798824399,0.124104000835143,0.0582935673014051 "150","SEC13","SEC13 homolog (S. cerevisiae)","protein","20100000 +/- 3.4e+07","65200000 +/- 7e+07",3.24,0.0039789499634028,0.0361722723945709,0.0172632536263772 "151","TPPP3","tubulin polymerization-promoting protein family member 3","protein","7290000 +/- 1.4e+07","41500000 +/- 1.1e+08",5.69,0.104161112556844,0.245084970721987,0.116353003680353 "152","TUBB1","tubulin, beta 1 class VI","protein","2.23e+08 +/- 2.3e+08","3.67e+08 +/- 4.5e+08",1.65,0.142154797337148,0.30099559436685,0.14311779504771 "153","PDLIM7","PDZ and LIM domain 7 (enigma)","protein","74800000 +/- 6.1e+07","1.24e+08 +/- 2.5e+08",1.66,0.322412476757539,0.495919508083587,0.235628047305912 "154","AGO1","argonaute RISC catalytic component 1","protein","4770000 +/- 1.6e+07","4880000 +/- 7500000",1.02,0.976205125939598,0.986065783777372,0.47124641934114 "155","UBQLN4","ubiquilin 4","protein","3640000 +/- 1.4e+07","6290000 +/- 1.1e+07",1.73,0.443755508306818,0.603748990893631,0.288323499649177 "156","ITGA11","integrin, alpha 11","protein","1210000 +/- 5800000","2520000 +/- 8800000",2.09,0.518136460203903,0.643647776650811,0.321131781932626 "157","SERPINA7","serpin peptidase inhibitor, clade A (alpha-1 antiproteinase, antitrypsin), member 7","protein","5810000 +/- 8400000","5230000 +/- 1.1e+07",0.9,0.832650934410059,0.871885795193779,0.431876394209847 "158","CDK5RAP3","CDK5 regulatory subunit associated protein 3","protein","630000 +/- 2300000","1780000 +/- 3e+06",2.83,0.117414150306826,0.263852023161407,0.125923645215935 "159","PTGR2","prostaglandin reductase 2","protein","244000 +/- 840000","1970000 +/- 4600000",8.09,0.0584244912949333,0.169178754049596,0.0798020813087872 "160","EIF5A2","eukaryotic translation initiation factor 5A2","protein","54300000 +/- 7.6e+07","1.47e+08 +/- 1.9e+08",2.71,0.0202493425401239,0.0987772806835311,0.0465881184407656 "161","PSMD9","proteasome (prosome, macropain) 26S subunit, non-ATPase, 9","protein","5810000 +/- 1.4e+07","53100000 +/- 8.5e+07",9.15,0.00626073521190854,0.0521727934325711,0.0238827121997089 "162","ZW10","zw10 kinetochore protein","protein","1340000 +/- 2100000","8930000 +/- 3e+07",6.68,0.198806126705078,0.386031313990442,0.177181208532096 "163","KDM1A","lysine (K)-specific demethylase 1A","protein","204000 +/- 820000","489000 +/- 1400000",2.4,0.368802231815649,0.530650693259926,0.253253580666222 "164","REXO2","RNA exonuclease 2","protein","1140000 +/- 2400000","6780000 +/- 9400000",5.95,0.00392053842648044,0.0361722723945709,0.0171607380039347 "165","ERMP1","endoplasmic reticulum metallopeptidase 1","protein","1730000 +/- 4100000","2670000 +/- 6300000",1.54,0.517582740222708,0.643647776650811,0.320898724006608 "166","TMEM214","transmembrane protein 214","protein","2540000 +/- 5500000","7630000 +/- 2.1e+07",3.01,0.224018049584567,0.40320029163737,0.189959771820575 "167","PGK2","phosphoglycerate kinase 2","protein","2.17e+08 +/- 3.7e+08","6.67e+08 +/- 8.7e+08",3.08,0.0169915871846209,0.0894294062348471,0.0426802746769151 "168","SEC63","SEC63 homolog (S. cerevisiae)","protein","1220000 +/- 2900000","3e+06 +/- 3900000",2.47,0.0615434310577959,0.173361777627594,0.0823138220429128 "169","COG4","component of oligomeric golgi complex 4","protein","182000 +/- 870000","545000 +/- 2e+06",3,0.390745104820936,0.554248375632533,0.263407465561433 "170","BPHL","biphenyl hydrolase-like (serine hydrolase)","protein","956000 +/- 3500000","1680000 +/- 3600000",1.76,0.457199778155707,0.613690977390211,0.294486312064147 "171","MYZAP","myocardial zonula adherens protein","protein","1170000 +/- 2200000","404000 +/- 1100000",0.34,0.109209836856161,0.25105709622106,0.11981725337821 "172","SCAF4","SR-related CTD-associated factor 4","protein","6450000 +/- 1.7e+07","11600000 +/- 1.7e+07",1.79,0.283915703124914,0.465435578893302,0.219184240026175 "173","PARP10","poly (ADP-ribose) polymerase family, member 10","protein","0 +/- 0","913000 +/- 4300000",Inf,0.27738906810875,0.458494327452479,0.216210927736388 "174","PPIL1","peptidylprolyl isomerase (cyclophilin)-like 1","protein","2710000 +/- 1e+07","4740000 +/- 8e+06",1.75,0.426736678632089,0.584570792646698,0.280365981074687 "175","CPSF3","cleavage and polyadenylation specific factor 3, 73kDa","protein","1050000 +/- 3400000","3630000 +/- 6700000",3.45,0.081050035926648,0.213289568228021,0.0991720420151246 "176","DHX38","DEAH (Asp-Glu-Ala-His) box polypeptide 38","protein","1130000 +/- 2200000","2280000 +/- 5100000",2.02,0.288725813352769,0.469472867240275,0.221339457787999 "177","FYN","FYN proto-oncogene, Src family tyrosine kinase","protein","31400000 +/- 6.7e+07","7880000 +/- 9700000",0.25,0.0768210588613783,0.204856156963675,0.0957130535770883 "178","AP3B2","adaptor-related protein complex 3, beta 2 subunit","protein","206000 +/- 650000","1450000 +/- 5200000",7.03,0.221232390852693,0.40320029163737,0.188609161250918 "179","MRTO4","mRNA turnover 4 homolog (S. cerevisiae)","protein","425000 +/- 1600000","6200000 +/- 2e+07",14.61,0.142972907324254,0.30099559436685,0.143650452183286 "180","DUT","deoxyuridine triphosphatase","protein","1860000 +/- 5100000","8170000 +/- 1.6e+07",4.4,0.0560335429212322,0.169178754049596,0.0778031581468442 "181","PDCD6","programmed cell death 6","protein","18700000 +/- 2.1e+07","49100000 +/- 6.8e+07",2.63,0.0299781343528746,0.119912537411498,0.0551326589686327 "182","LHPP","phospholysine phosphohistidine inorganic pyrophosphate phosphatase","protein","8140000 +/- 1.1e+07","38200000 +/- 5.2e+07",4.7,0.00464695818720693,0.0404083320626689,0.0192849173921797 "183","ARMCX3","armadillo repeat containing, X-linked 3","protein","20200 +/- 110000","2830000 +/- 8900000",139.8,0.105890385832605,0.246256711238617,0.117526264784731 "184","WDR33","WD repeat domain 33","protein","0 +/- 0","297000 +/- 1500000",Inf,0.321941348172834,0.495919508083587,0.235437415083026 "185","CPA6","carboxypeptidase A6","protein","3310000 +/- 1.4e+07","2220000 +/- 1.2e+07",0.67,0.760472260059229,0.831117224108447,0.40978011063697 "186","WDR82","WD repeat domain 82","protein","317000 +/- 1200000","1410000 +/- 3600000",4.45,0.140038375407462,0.30099559436685,0.141729764384562 "187","IGFN1","immunoglobulin-like and fibronectin type III domain containing 1","protein","231000 +/- 850000","110000 +/- 460000",0.47,0.517772301943548,0.643647776650811,0.320978527583109 "188","MEMO1","mediator of cell motility 1","protein","186000 +/- 950000","1900000 +/- 5e+06",10.24,0.0870740656603324,0.217685164150831,0.103890465007097 "189","HRSP12","heat-responsive protein 12","protein","33400 +/- 170000","1040000 +/- 4100000",31.2,0.205620752278862,0.391658575769262,0.180762004992734 "190","NT5DC1","5'-nucleotidase domain containing 1","protein","280000 +/- 810000","3240000 +/- 8e+06",11.55,0.0628775387710606,0.174659829919613,0.0833565806801436 "191","TRIM21","tripartite motif containing 21","protein","129000 +/- 670000","4e+06 +/- 2e+07",30.99,0.308827237385239,0.495919508083587,0.230029273261299 "192","PTPRK","protein tyrosine phosphatase, receptor type, K","protein","701000 +/- 3e+06","162000 +/- 680000",0.23,0.36553148555418,0.530650693259926,0.252080934181609 "193","KDM3B","lysine (K)-specific demethylase 3B","protein","1420000 +/- 7200000","148000 +/- 440000",0.1,0.365428424559181,0.530650693259926,0.252043821164736 "194","BNIP1","BCL2/adenovirus E1B 19kDa interacting protein 1","protein","108000 +/- 560000","177000 +/- 920000",1.64,0.739341569549298,0.81695201055171,0.402982310572385 "195","PCLO","piccolo presynaptic cytomatrix protein","protein","742000 +/- 2800000","235000 +/- 710000",0.32,0.367711772521732,0.530650693259926,0.25286373427761 "196","F13B","coagulation factor XIII, B polypeptide","protein","25300 +/- 130000","0 +/- 0",0,0.321941348172834,0.495919508083587,0.235437415083026 "197","ANKRD28","ankyrin repeat domain 28","protein","48800 +/- 190000","462000 +/- 950000",9.46,0.031989677210485,0.120715763058434,0.0564886159975311 "198","MATN3","matrilin 3","protein","0 +/- 0","23900 +/- 120000",Inf,0.321941348172835,0.495919508083587,0.235437415083026 "199","USP34","ubiquitin specific peptidase 34","protein","2980000 +/- 9300000","149000 +/- 770000",0.05,0.122564845706434,0.272366323792075,0.129685739762852 "200","DCAF7","DDB1 and CUL4 associated factor 7","protein","599000 +/- 3e+06","1320000 +/- 2800000",2.21,0.36265855374751,0.530650693259926,0.251042609848482 ================================================ FILE: Demos/Data Analysis Workflow/results/statistical_results_sig.csv ================================================ ,ID,description,type,healthy.mean.....std.dev,sick.mean.....std.dev,mean.sick_mean.healthy,group_p.values,group_adjusted.p.values,group_q.values 35,C00077,ornithine,metabolite,7300 +/- 2900,3910 +/- 2100,0.54,9.43E-06,0.001886663,0.000900412 12,C02477,tocopherol alpha,metabolite,1160 +/- 1400,2600 +/- 1200,2.24,0.000150503,0.007714558,0.003632462 74,C00097,cysteine,metabolite,5160 +/- 2500,8460 +/- 3400,1.64,0.000150784,0.007714558,0.003633836 65,C00031,glucose,metabolite,335000 +/- 2e+05,144000 +/- 140000,0.43,0.000154291,0.007714558,0.003650646 92,C00170,5'-deoxy-5'-methylthioadenosine,metabolite,216 +/- 83,355 +/- 160,1.64,0.000200406,0.008016221,0.003825749 2,C00385,xanthine,metabolite,574 +/- 350,1670 +/- 1500,2.91,0.000390905,0.013030159,0.006072846 77,C00158,citric acid,metabolite,3640 +/- 2000,1930 +/- 1400,0.53,0.000579756,0.016564459,0.007603203 17,D09007,tagatose,metabolite,163 +/- 61,108 +/- 56,0.66,0.00110566,0.023396521,0.01011206 63,C00025,glutamic acid,metabolite,72900 +/- 18000,96200 +/- 30000,1.32,0.001112529,0.023396521,0.010134825 129,IVD,isovaleryl-CoA dehydrogenase,protein,1.1e+07 +/- 1.4e+07,58900000 +/- 7.1e+07,5.34,0.001169826,0.023396521,0.010317932 134,GLOD4,glyoxalase domain containing 4,protein,24500000 +/- 4.5e+07,1.24e+08 +/- 1.4e+08,5.06,0.001296961,0.023581103,0.010685382 76,C00327,citrulline,metabolite,597 +/- 240,401 +/- 180,0.67,0.001629237,0.024873585,0.011450641 69,C01018,fucose+ rhamnose,metabolite,1150 +/- 730,2500 +/- 2000,2.17,0.001754997,0.024873585,0.011684702 81,C06555,biuret,metabolite,269 +/- 180,150 +/- 66,0.56,0.001855321,0.024873585,0.011854452 95,C00197,3-phosphoglycerate,metabolite,1030 +/- 660,510 +/- 490,0.49,0.001865519,0.024873585,0.011870943 75,C00791,creatinine,metabolite,10800 +/- 4200,15500 +/- 6600,1.44,0.002880815,0.034787816,0.014974523 94,C00989,4-hydroxybutyric acid,metabolite,209 +/- 42,282 +/- 110,1.35,0.002956964,0.034787816,0.015162094 1,C00379,xylitol,metabolite,579 +/- 300,880 +/- 410,1.52,0.003285451,0.036172272,0.01591618 19,C00315,spermidine,metabolite,276 +/- 100,189 +/- 110,0.68,0.003764835,0.036172272,0.016878367 144,TXNL1,thioredoxin-like 1,protein,12400000 +/- 2.2e+07,46100000 +/- 5.4e+07,3.73,0.003809829,0.036172272,0.01696136 164,REXO2,RNA exonuclease 2,protein,1140000 +/- 2400000,6780000 +/- 9400000,5.95,0.003920538,0.036172272,0.017160738 150,SEC13,SEC13 homolog (S. cerevisiae),protein,20100000 +/- 3.4e+07,65200000 +/- 7e+07,3.24,0.00397895,0.036172272,0.017263254 182,LHPP,phospholysine phosphohistidine inorganic pyrophosphate phosphatase,protein,8140000 +/- 1.1e+07,38200000 +/- 5.2e+07,4.7,0.004646958,0.040408332,0.019284917 161,PSMD9,"proteasome (prosome, macropain) 26S subunit, non-ATPase, 9",protein,5810000 +/- 1.4e+07,53100000 +/- 8.5e+07,9.15,0.006260735,0.052172793,0.023882712 89,C00020,adenosine-5-phosphate,metabolite,1300 +/- 1500,3020 +/- 2800,2.32,0.006837402,0.054699215,0.02535055 79,C06423,caprylic acid,metabolite,1470 +/- 530,1080 +/- 490,0.74,0.007348549,0.056527301,0.026584412 126,HNRNPF,heterogeneous nuclear ribonucleoprotein F,protein,1.59e+08 +/- 2.4e+08,4.12e+08 +/- 4.2e+08,2.59,0.00853303,0.06320763,0.029225688 43,C00047,lysine,metabolite,24900 +/- 25000,10700 +/- 12000,0.43,0.009961495,0.071153537,0.032059435 66,C00257,gluconic acid,metabolite,502 +/- 580,199 +/- 120,0.4,0.010517694,0.072535824,0.033072427 145,EPS8,epidermal growth factor receptor pathway substrate 8,protein,411000 +/- 1300000,6240000 +/- 1.1e+07,15.2,0.011064938,0.073766254,0.03402472 50,C00130,inosine 5'-monophosphate,metabolite,199 +/- 340,625 +/- 790,3.14,0.013056504,0.081037107,0.037161668 104,XRCC6,X-ray repair complementing defective repair in Chinese hamster cells 6,protein,2.51e+08 +/- 3.4e+08,7.33e+08 +/- 9.1e+08,2.92,0.013079477,0.081037107,0.037195132 28,C00346,phosphoethanolamine,metabolite,13500 +/- 7700,21200 +/- 14000,1.57,0.01341153,0.081037107,0.037672494 112,EEF1D,eukaryotic translation elongation factor 1 delta (guanine nucleotide exchange protein),protein,2.12e+08 +/- 2.6e+08,5.18e+08 +/- 5.7e+08,2.45,0.013776308,0.081037107,0.038183608 16,C00245,taurine,metabolite,93000 +/- 29000,125000 +/- 59000,1.34,0.014225218,0.081286961,0.038794284 56,C00135,histidine,metabolite,6550 +/- 3400,4720 +/- 1700,0.72,0.015799645,0.087775805,0.041083803 48,C03546,inositol-4-monophosphate,metabolite,302 +/- 130,218 +/- 120,0.72,0.016408041,0.088692114,0.041912481 167,PGK2,phosphoglycerate kinase 2,protein,2.17e+08 +/- 3.7e+08,6.67e+08 +/- 8.7e+08,3.08,0.016991587,0.089429406,0.042680275 119,ADH5,"alcohol dehydrogenase 5 (class III), chi polypeptide",protein,1.18e+08 +/- 1.1e+08,2.33e+08 +/- 2.2e+08,1.98,0.018476135,0.094749413,0.044546442 114,NAPSA,napsin A aspartic peptidase,protein,3.33e+08 +/- 3.2e+08,1.37e+09 +/- 2.2e+09,4.11,0.020058357,0.098777281,0.046377239 160,EIF5A2,eukaryotic translation initiation factor 5A2,protein,54300000 +/- 7.6e+07,1.47e+08 +/- 1.9e+08,2.71,0.020249343,0.098777281,0.046588118 30,C01601,pelargonic acid,metabolite,18800 +/- 8300,13900 +/- 6700,0.74,0.021076289,0.10036328,0.047477726 128,SEC24C,SEC24 family member C,protein,14500000 +/- 2.2e+07,38500000 +/- 4.8e+07,2.66,0.022339763,0.103905873,0.048767383 132,API5,apoptosis inhibitor 5,protein,16700000 +/- 3.2e+07,39900000 +/- 4.2e+07,2.39,0.025451741,0.113425849,0.051627698 113,GLS,glutaminase,protein,13900000 +/- 3.1e+07,83700000 +/- 1.5e+08,6.04,0.025520816,0.113425849,0.051686601 61,C00116,glycerol,metabolite,22900 +/- 7600,27700 +/- 8200,1.21,0.027020787,0.116240884,0.052921588 133,RPL9,ribosomal protein L9,protein,96200000 +/- 1e+08,1.88e+08 +/- 1.8e+08,1.96,0.027316608,0.116240884,0.053155604 87,C00041,alanine,metabolite,213000 +/- 59000,258000 +/- 86000,1.21,0.029684811,0.119912537,0.054925477 88,C06104,adipic acid,metabolite,926 +/- 680,591 +/- 370,0.64,0.029795109,0.119912537,0.055003679 181,PDCD6,programmed cell death 6,protein,18700000 +/- 2.1e+07,49100000 +/- 6.8e+07,2.63,0.029978134,0.119912537,0.055132659 26,C00148,proline,metabolite,68200 +/- 38000,94500 +/- 49000,1.39,0.031161063,0.120715763,0.05594336 131,ACSS2,acyl-CoA synthetase short-chain family member 2,protein,698000 +/- 2e+06,12700000 +/- 2.8e+07,18.25,0.03183592,0.120715763,0.056388794 197,ANKRD28,ankyrin repeat domain 28,protein,48800 +/- 190000,462000 +/- 950000,9.46,0.031989677,0.120715763,0.056488616 7,C00106,uracil,metabolite,16300 +/- 11000,24700 +/- 17000,1.52,0.033473358,0.123975402,0.057421501 149,EVL,Enah/Vasp-like,protein,6110000 +/- 8600000,21900000 +/- 3.7e+07,3.59,0.03494138,0.124104001,0.058293567 135,ABHD10,abhydrolase domain containing 10,protein,2530000 +/- 6700000,20100000 +/- 4.2e+07,7.93,0.03515291,0.124104001,0.058415292 72,C00836,dihydrosphingosine,metabolite,341 +/- 150,234 +/- 210,0.69,0.036002611,0.124104001,0.058894733 80,C01571,capric acid,metabolite,2800 +/- 830,2240 +/- 1100,0.8,0.036375904,0.124104001,0.059100669 5,C00366,uric acid,metabolite,2960 +/- 2100,4150 +/- 1900,1.4,0.03661068,0.124104001,0.05922876 40,C01835,maltotriose,metabolite,3290 +/- 2200,7370 +/- 9700,2.24,0.037662655,0.125542182,0.059915134 123,SLFN5,schlafen family member 5,protein,10100000 +/- 1.4e+07,2.1e+07 +/- 2.3e+07,2.07,0.042904472,0.140233853,0.065529479 118,ANXA7,annexin A7,protein,81800000 +/- 7.5e+07,1.54e+08 +/- 1.7e+08,1.88,0.043472494,0.140233853,0.066111069 82,C00180,benzoic acid,metabolite,12200 +/- 3100,10000 +/- 4400,0.83,0.04685931,0.148759716,0.069477396 ================================================ FILE: Demos/Mapped Network From Data (Biochemical and Structural)/data.csv ================================================ ID,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92 1,0,0.000234,0,0.00329,0,0.367,0,0.000706,0.000564,0.0874,0.000425,0.00094,0,0.000645,0.000178,0.000341,0,0,0,0.00072,0.00463,0.00276,0,0,0.00339,0,0.00228,0,0,0.000324,0.000102,0.00196,0.000937,0,0,0,0.000153,0.000778,0.00069,0.00573,0.000863,0,0.00169,0,0.000237,0.000306,0,0,0,0.0504,0,0.021,0.000732,0.00177,0,0.00292,0.0017,0,0.00107,0.00816,0.0509,0.0065,0.00309,0,0,0.00275,0.00387,0,0.00423,0.00128,0,0,0.0023,0.056,0.000795,0,0.00049,0.000767,0.00154,0.00143,0,0,0.00272,0.00849,0.000809,0.00101,0,0.00077,0,0,0,0.000241 2,0,0.000228,0,0.00388,0,0.213,0,0.000698,0.000952,0.0542,0.000268,0.00417,0,0.000308,0,0.000227,0.0000927,0,0,0.0009,0.00521,0.00207,0,0,0.00519,0,0.00481,0,0,0.000108,0,0.000399,0.00103,0.000395,0,0,0,0.00392,0.000599,0.00697,0.00123,0,0.00183,0,0,0.000112,0,0,0.0000789,0.00621,0,0.0314,0,0,0,0.001597,0.00753,0,0.00269,0.0055,0.00509,0.00713,0.00424,0,0,0.00144,0.00464,0,0.00324,0.00431,0,0,0.0341,0.118,0,0.00119,0.0000928,0,0.00324,0.0013,0,0,0.00401,0.00993,0.000777,0.000544,0.000745,0.000755,0,0.00292,0.009,0 3,0,0,0,0,0,0,0,0,0,0.0579,0,0,0,0,0,0.000143,0,0,0,0.00702,0.0506,0,0.000369,0.000273,0.00102,0,0,0,0,0,0,0.00297,0,0,0.00253,0,0,0,0.0015,0.054,0.00277,0.00183,0,0.0521,0.000407,0,0,0,0,0.14,0,0,0.00286,0,0,0,0.00596,0,0,0.0267,0,0.0667,0,0,0.0718,0,0,0,0,0.00306,0,0.00136,0.00503,0,0,0,0,0,0.000531,0,0,0,0.00227,0,0.00113,0,0,0,0.00102,0.00591,0,0.000539 4,0,0.000286,0.00021,0.00254,0,0.241,0,0.001324,0.000476,0.0558,0.000661,0.00354,0,0.000458,0.000119,0.000327,0,0.0005,0,0.00132,0.00449,0.00364,0,0,0.00462,0,0.00247,0.0009,0,0.000558,0,0.000703,0.00112,0,0,0,0,0.00222,0.0005,0.00655,0.0014,0.00083,0.00121,0,0.000141,0.000566,0,0,0.0000752,0.0205,0,0.0522,0.000147,0,0,0.0055,0,0.000943,0,0.00385,0.0075,0.0351,0.00477,0,0,0.00145,0.00397,0,0.00585,0.00271,0,0,0.0068,0.0681,0,0.000414,0.0013,0,0.00184,0.00137,0,0,0.00379,0.0521,0.00161,0.00523,0.00208,0.0563,0.000546,0.00185,0,0.000243 5,0,0,0,0.00492,0,0.183,0,0.000236,0.000658,0.0438,0,0.00108,0,0,0.0000805,0.000417,0,0,0,0.00102,0.004,0.00128,0,0,0.00386,0,0.00131,0,0,0.0000882,0,0.000234,0.000869,0,0,0,0,0,0.00047,0.00556,0.000703,0,0.00137,0,0.000349,0.000115,0,0,0.000112,0.0678,0,0.0436,0.00026,0.000683,0.000443,0.001761,0,0.00562,0.00519,0.00117,0.00086,0,0.00759,0.00436,0,0.00367,0.00178,0.0039,0.00338,0.00317,0,0.0000806,0.00189,0.0808,0.000728,0.000766,0.0000743,0,0.00375,0.00057,0,0,0.00111,0.00871,0.00107,0.000838,0,0,0.000582,0.0508,0,0.005793 6,0,0.00054,0.00014,0.00228,0,0.177,0,0.0014,0.000678,0.0574,0.000538,0.00586,0,0.000423,0.000121,0.000387,0.0000662,0.000113,0,0.00121,0.00543,0.00338,0,0,0.00411,0.00015,0.00225,0,0,0.000444,0.0000665,0.000744,0.00128,0,0,0.0005,0.000155,0.05946,0.00096,0.00901,0.00161,0.00115,0.00149,0.00107,0.000167,0.000253,0.00044,0.00119,0.0001,0.0512,0.00336,0.0298,0.000509,0.000613,0,0.00452,0.00704,0,0.0045,0.00637,0.0552,0.00708,0.00639,0,0,0.00218,0.00224,0,0.00292,0.00258,0.0000765,0.000262,0.00391,0.0906,0.00162,0.000417,0.00116,0,0.00851,0.00192,0,0.0504,0.00649,0.0547,0.00123,0.00488,0.000299,0.00287,0.000413,0.00218,0,0.000247 7,0,0,0,0.00449,0,0.281,0,0.000187,0.000644,0.0375,0,0.00359,0,0,0,0,0,0,0,0.000595,0.00345,0.00148,0,0,0.00287,0,0.00232,0,0,0,0,0.000495,0,0,0,0,0,0.000916,0.0004,0.00457,0.00106,0,0.00108,0,0,0,0,0,0,0.00317,0,0.0511,0,0.000707,0,0.000846,0.00133,0,0.00341,0.00318,0,0.00228,0.00296,0,0,0.00752,0.00113,0,0.0053,0,0,0,0.00211,0.00373,0,0,0,0,0.00381,0.000503,0,0,0.00398,0.0081,0.00108,0.000849,0,0,0,0.0584,0,0.00293 8,0,0,0,0.05,0,0.297,0,0.000199,0,0.0284,0.000146,0,0,0.000291,0.000254,0.000668,0,0,0,0.00147,0.00987,0.00248,0.000309,0,0.00594,0,0,0,0,0,0.000273,0.00122,0.00164,0,0,0,0,0,0.00088,0.0508,0.00263,0.00221,0.00158,0,0.000535,0,0,0.00117,0.000261,0.00364,0,0.0339,0.00167,0,0,0.000781,0.00955,0,0,0.05576,0,0.0053,0.00578,0,0,0.00259,0,0.00102,0.0083,0.0014,0,0,0.00262,0.0524,0.00119,0,0,0,0.00696,0.00149,0,0,0.00369,0.009,0.00162,0.0014,0,0,0.00142,0.0041,0,0 9,0,0,0,0.00671,0,0.266,0,0.000531,0.0015,0.0212,0,0.00198,0,0.000398,0.000177,0.00046,0.00314,0,0,0.00058,0.00657,0.00239,0.0000936,0,0.00623,0,0.00224,0,0.00109,0.000138,0,0.000432,0.00131,0.00127,0.000815,0,0,0,0.00109,0.00857,0.00141,0,0.00224,0,0.000396,0,0,0,0.00159,0.0256,0,0.0532,0.00905,0,0,0.000789,0.00169,0,0,0.00587,0.00139,0.00502,0.00482,0,0,0,0.00235,0,0.00571,0.00363,0,0.000174,0.00314,0.0758,0.00378,0.000707,0.000145,0,0.00611,0.00123,0,0,0,0.0572,0.00183,0.00227,0.00075,0.00222,0.000501,0,0,0 10,0,0,0,0.0212,0,0.136,0.00164,0,0,0.0507,0,0,0.000296,0.000395,0.000226,0.000678,0,0,0,0.00505,0.0503,0.00217,0.000537,0,0.00674,0,0,0,0,0,0.000349,0.00129,0.0027,0,0.00284,0,0.000545,0,0.00101,0.0592,0.00298,0.00288,0.00155,0,0.00125,0,0,0,0.000319,0.0207,0,0.0551,0.00736,0,0,0.0013,0,0.00128,0.05677,0.05803,0,0.00374,0.00719,0,0,0.00363,0,0,0.00873,0.00297,0,0.000262,0.00498,0.00225,0.00517,0,0,0,0.00665,0.00456,0,0,0.00328,0.0537,0.00386,0.00258,0,0,0.00116,0.00511,0,0.00201 11,0,0.000483,0,0.00701,0,0.231,0,0.000437,0.000907,0.0327,0,0.00343,0,0.000355,0.000087,0.000442,0.000338,0.00021,0,0.0006,0.00441,0.0018,0,0,0.00392,0,0.00319,0,0,0,0,0.00234,0.000939,0,0.000573,0,0,0,0.000934,0.0074,0.00155,0.00146,0.00143,0.000569,0,0.000214,0,0.00118,0.000141,0.00277,0,0.0389,0,0,0,0.00242,0.00432,0,0,0.00685,0.00106,0.00504,0.00526,0,0,0,0.00435,0,0.00672,0.00239,0,0,0.00692,0.0419,0,0.00118,0,0,0.00824,0.000535,0,0,0.0041,0.0537,0.00188,0.00225,0,0.000726,0.000339,0.000372,0,0 12,0,0.000315,0,0.00109,0,0.173,0,0.0005,0.000921,0.0266,0,0.00326,0,0.000358,0,0.000246,0,0,0,0.000383,0.00344,0.00137,0,0,0.0032,0,0.00506,0,0,0.000364,0,0.00179,0.00189,0,0,0,0,0.00387,0.00088,0.00666,0.00111,0,0.001,0,0,0.000201,0.000573,0,0.000288,0.00229,0,0.0336,0,0,0,0.00454,0.00408,0,0,0.00648,0.00606,0.00532,0.00506,0,0,0.00152,0.00565,0,0.00459,0.0029,0.0000784,0,0.00734,0.102,0,0.00121,0.0012,0,0.00835,0.000508,0,0,0.00371,0.0541,0.00165,0.00152,0.00101,0.00884,0.000393,0.000716,0,0 13,0,0,0,0.00841,0,0.0295,0,0,0,0.0462,0.000246,0,0,0.000601,0.000277,0.00155,0,0,0,0.00335,0.0576,0.00303,0.000504,0,0.0502,0,0,0,0.00226,0,0.000314,0.00875,0.00641,0,0.00167,0,0,0,0.00171,0.0582,0.00416,0.00921,0.00486,0,0.00122,0.000166,0,0.000449,0,0.0248,0,0.0356,0.00245,0.00205,0,0,0.0558,0,0,0.02941,0,0.148,0.0539,0,0,0.00318,0,0.0025,0.0515,0.00416,0,0.000794,0.00478,0,0.0028,0.00277,0,0.00138,0.00949,0.00254,0,0,0.00547,0.000883,0.00399,0.000512,0.000216,0,0.00314,0.00171,0.00575,0.000631 14,0,0.000302,0.000181,0.00212,0,0.222,0,0.00137,0.000826,0.0238,0.000604,0.00538,0,0.000152,0.000243,0.000595,0.000266,0,0,0.00127,0.00603,0.00183,0,0,0.00583,0,0.00219,0,0.000958,0.000421,0,0.000393,0.0015,0.00246,0.000647,0,0,0.00488,0.000526,0.00652,0.00135,0.000697,0.00202,0.000357,0.000206,0,0.00198,0,0.000313,0.00477,0,0.0819,0,0.00115,0.000439,0.003083,0,0.00742,0,0.00715,0.00957,0.00396,0.00731,0,0,0.00178,0.00339,0,0.00303,0.00096,0,0.00026,0.00166,0.0793,0,0.000657,0.000961,0.00352,0.00678,0.000456,0,0,0.00375,0.0504,0.00104,0.00485,0,0.000361,0.000814,0.00599,0,0.0000897 15,0,0,0.0000798,0.00332,0,0.273,0,0.000419,0.000531,0.0459,0.000136,0.000911,0,0.000133,0.000268,0.00038,0.000264,0,0,0.00157,0.0075,0.00176,0,0,0.00713,0,0.00394,0,0,0.0000941,0.000116,0.000672,0.00247,0.000427,0.000421,0,0.000321,0.000334,0.000718,0.00832,0.00104,0,0.0016,0,0.0000729,0.0000907,0.000314,0,0.00104,0.00348,0,0.034,0,0,0,0.000736,0.0021,0.00242,0,0.004688,0.000967,0.00471,0.00549,0,0.00708,0.00192,0.00236,0,0.00396,0.00166,0,0,0.0019,0.101,0,0.00116,0.000241,0,0.00512,0.000709,0,0,0.00336,0.0507,0.00119,0.00167,0,0.00118,0.0007,0.004,0,0.00113 16,0,0,0,0.00319,0,0.156,0,0.000155,0.000428,0.0248,0,0.00121,0,0.000183,0,0.000244,0.00734,0,0,0.000671,0.00448,0.00155,0,0,0.00394,0,0.00306,0,0,0.000111,0.000869,0,0.00127,0,0,0,0,0,0,0.0061,0.00116,0.00103,0.00102,0,0.00013,0,0,0,0.000225,0.00288,0,0.032,0.000618,0.000791,0,0.000988,0.00398,0,0,0.00878,0.00146,0.00491,0.00393,0,0,0.00155,0.00245,0,0.00784,0.00237,0,0,0.00551,0.186,0.000721,0.00129,0,0,0.00742,0.000664,0,0,0.00273,0.0568,0.00207,0.000562,0,0,0.000946,0.00549,0,0.000246 17,0,0.000391,0.000237,0.0017,0.0000597,0.208,0,0.00265,0.000712,0.0383,0.002494,0.00818,0,0.000521,0.0000546,0.000255,0.00063,0.000364,0,0.000366,0.00331,0.00363,0,0,0.00228,0,0.00372,0.00049,0.000628,0.000571,0,0.000737,0.00117,0.00194,0,0,0,0.00381,0.0006,0.00512,0.00112,0,0.00114,0,0,0.000144,0,0,0,0.0036,0,0.0859,0,0.000803,0,0.00446,0,0,0,0.00847,0.0579,0.171,0.00407,0,0,0,0.00495,0.000375,0.00506,0.0017,0,0.000103,0.00375,0.00701,0,0,0.00107,0,0.00165,0.00043,0,0.000327,0.001,0.00923,0.000939,0.000671,0.000462,0.00531,0.000287,0,0,0 18,0,0,0,0.00656,0,0.191,0,0.0000732,0.000197,0.0435,0,0.00131,0,0,0,0,0,0,0,0.0009,0.00239,0.00137,0,0,0.00154,0,0.000693,0,0,0,0,0.000975,0.000579,0,0,0,0,0,0.0001,0.00312,0.000558,0,0.000506,0.000372,0,0,0,0,0,0.0501,0,0.0398,0.000597,0,0,0.000293,0.00274,0,0,0.00479,0,0.279,0.00294,0,0,0,0.00129,0,0.00275,0.00138,0,0,0.00407,0.053,0,0,0,0,0.00189,0.00119,0,0,0.00444,0.00377,0.00079,0.000679,0.000197,0,0,0,0,0 19,0,0.000283,0,0.00442,0,0.269,0,0.000139,0.000445,0.0546,0,0.00087,0,0.00023,0.00034,0.00104,0,0,0,0.0018,0.00611,0.00133,0,0,0.00433,0,0.00267,0,0,0,0,0.000492,0.00111,0,0,0,0,0,0.0005,0.00562,0.000933,0,0.00156,0.000419,0.000285,0,0,0,0,0.09172,0,0.0588,0.00246,0,0,0.00146,0.0523,0,0,0.00483,0.000635,0.0593,0.0035,0,0,0.00318,0.00106,0,0.00441,0.00355,0,0,0.000645,0.0306,0.00272,0.000373,0.000127,0,0.00324,0.000607,0,0,0.00217,0.00798,0.000924,0.000932,0,0.00483,0.00033,0.00771,0,0.010296 20,0,0,0,0.00968,0,0.359,0,0,0.000516,0.0559,0,0.000496,0,0,0,0,0,0,0,0,0.00262,0.00119,0,0,0.00198,0,0.000799,0,0,0,0,0.000779,0,0,0,0,0,0,0,0.00412,0,0,0,0.000803,0,0,0,0,0,0.00351,0,0.0531,0,0,0,0,0.00226,0,0,0.00263,0,0.00388,0.00171,0,0,0,0.00129,0,0.00289,0.00149,0,0,0.00161,0.00227,0,0,0,0,0.0013,0.00025,0,0,0.00231,0.00428,0.000701,0,0.00115,0,0,0,0,0 21,0,0.000167,0,0.00618,0,0.174,0,0.000389,0.000869,0.028,0,0.00164,0,0,0.000208,0.000969,0,0,0,0.00111,0.00771,0.00207,0.000168,0,0.00586,0,0.0022,0,0.00109,0.000117,0.0000989,0.00158,0.00202,0,0.000704,0,0,0,0.000629,0.00975,0.00122,0,0.00236,0.000549,0.00056,0,0,0,0,0.00865,0,0.102,0.00196,0,0,0.001028,0,0.000847,0.0015,0.05178,0,0.0386,0.00766,0,0,0.00193,0.00169,0,0.00445,0.00831,0,0,0.0299,0.0513,0.00309,0,0,0,0.00909,0.00115,0,0,0.00612,0.0527,0.00164,0.00066,0.000581,0,0.00111,0.0521,0,0.00143 22,0,0,0,0.00413,0,0.463,0,0.000206,0.000892,0.0452,0,0.00131,0,0.000309,0,0,0,0,0,0.0004,0.00282,0.000954,0,0,0.00204,0,0.00888,0,0,0.000131,0,0.000432,0,0.00086,0,0,0,0,0,0.00257,0.000128,0,0.000378,0,0,0,0,0,0,0.00521,0,0.0201,0.00086,0,0,0.001016,0,0,0,0.00135,0.0022,0.00228,0.000765,0,0,0,0.00368,0,0.0012,0.00151,0,0,0.00047,0.0226,0.000601,0,0.0000855,0,0.00102,0,0,0,0.000597,0.00278,0.000312,0,0.000172,0,0,0,0,0 23,0,0,0,0.00127,0,0.23,0,0.000657,0.000529,0.0414,0.000125,0.00111,0,0.000164,0.000228,0.000324,0,0,0,0.0012,0.00445,0.00189,0,0,0.00351,0,0.00219,0,0,0.000212,0,0.000508,0.00149,0,0,0,0,0.00505,0.000458,0.0076,0.00166,0,0.00156,0,0.000122,0.000131,0,0,0,0.00779,0,0.0418,0.000279,0.00083,0,0.0028,0.00554,0,0.00335,0.00549,0.0057,0.0497,0.00373,0,0,0.0033,0.00171,0.00125,0.0518,0.00234,0,0.000239,0.0226,0.0883,0,0.000725,0.000504,0,0.00343,0.000701,0,0,0.00717,0.0519,0.00118,0.001,0.000223,0.00107,0.000299,0.00216,0,0.0018 24,0.000214,0,0,0.00406,0,0.182,0,0,0.000533,0.0376,0,0.000901,0,0,0,0.000284,0,0,0,0.001,0.00347,0.000911,0,0,0.00318,0,0.00136,0,0,0,0,0.000549,0.000639,0,0,0,0,0,0.000295,0.00221,0.000794,0,0.00115,0.000485,0,0,0,0,0,0.0535,0,0.0515,0.000943,0,0,0.000275,0.00115,0,0,0,0,0,0.00167,0,0,0.0527,0.000826,0,0,0.00473,0,0,0.00332,0.108,0.00159,0,0,0,0.0027,0,0,0,0,0.0506,0.00142,0,0.00104,0.00182,0,0.00263,0,0.00395 25,0,0,0,0.00179,0,0.281,0,0.000594,0.000846,0.0283,0.000259,0.00137,0,0.00103,0,0,0.000237,0,0,0.00046,0.00327,0.00184,0,0,0.00265,0,0.00467,0,0,0.000251,0,0.000243,0.00195,0.000277,0,0,0,0.00167,0,0.00506,0.000693,0,0.000569,0,0,0,0.000473,0,0.000562,0.00609,0,0.0244,0.000262,0.000353,0,0.00201,0.00747,0,0,0.003467,0.00291,0.00765,0.00267,0,0,0,0.00273,0,0.00276,0.00228,0,0,0.00643,0.144,0,0,0.000221,0,0.00149,0.000239,0,0,0.00189,0.0078,0.000882,0,0.000642,0.00374,0,0,0,0 26,0,0.000265,0,0.00803,0,0.204,0,0.000356,0.00123,0.0699,0.000236,0.000956,0,0.000326,0.0000747,0.000605,0,0,0,0.00043,0.00594,0.002,0,0,0.00519,0,0.00183,0.0008,0,0.0000862,0,0.000319,0.000673,0,0,0,0,0,0.0008,0.00722,0.00117,0.002264,0.00258,0,0.000055,0.000705,0,0,0.00016,0.00407,0,0.028,0.000692,0.000721,0,0.00256,0,0.0078,0,0.0097,0.00146,0.00774,0.00676,0,0,0.00227,0.00364,0,0.00424,0.00235,0,0.000144,0.0025,0.0628,0.000877,0.00267,0.000153,0,0.0516,0.00167,0,0,0.000671,0.0541,0.00208,0.00326,0,0,0.000565,0.04,0,0.00311 27,0,0.000154,0,0.00661,0,0.199,0,0.000747,0.00114,0.0707,0.000163,0.00272,0,0.00112,0.000113,0.000282,0.0000812,0,0,0.00352,0.0556,0.00259,0.0003,0.0014,0.00769,0,0.00423,0,0.00115,0.000196,0.000118,0.000841,0.00092,0.000546,0.000847,0,0,0.000487,0.000775,0.0503,0.00286,0.00165,0.00187,0,0.000349,0.000299,0,0,0.000261,0.00602,0,0.0615,0.00129,0.0011,0.000359,0.003775,0,0.00144,0,0.00737,0.00415,0.00906,0.00978,0,0,0.00155,0.00674,0,0.00412,0.00335,0.000161,0,0.00306,0.055,0.00378,0.00131,0.000243,0,0.00744,0.00542,0,0,0.0579,0.0566,0.00184,0.00265,0.000251,0.000777,0.000807,0.00143,0,0 28,0,0,0,0.00492,0,0.27,0,0.000201,0.00107,0.053,0,0.000926,0,0.000201,0.0000984,0,0.000145,0,0,0.000734,0.00428,0.00208,0,0,0.0046,0,0.00176,0,0.000992,0.000134,0,0.000446,0.00146,0,0,0,0,0,0.00065,0.00721,0.00115,0,0.00134,0,0.000136,0.000321,0,0,0.000169,0.00412,0,0.0358,0.000386,0.000436,0,0.00116,0.00249,0,0.00246,0.007122,0.00103,0.159,0.00398,0,0,0.00945,0.00255,0,0.0051,0.00159,0,0.000232,0.00194,0.00523,0.000487,0,0.0000907,0,0.00353,0.000541,0,0,0.0041,0.0503,0.00101,0.00384,0.00048,0.000781,0.000271,0.00123,0,0.0018 29,0,0.000354,0,0.00658,0,0.171,0,0.000214,0.000763,0.0565,0,0.00316,0,0.000248,0.000269,0.000707,0.000206,0,0,0.0021,0.00574,0.00235,0.000179,0,0.00496,0,0.00179,0,0.000787,0,0.000194,0.000493,0.00157,0,0,0,0,0.000555,0.000534,0.00969,0.00172,0,0.00227,0.000575,0.00018,0.000205,0,0,0,0.0557,0,0.1,0.000821,0.000987,0,0.000667,0.0219,0.00149,0,0.007512,0.00303,0.00973,0.00466,0,0,0.00475,0.00315,0,0.00601,0.00422,0,0,0.00764,0.0513,0.00117,0,0.000116,0,0.00866,0.000977,0,0,0.0054,0.0584,0.00226,0.00324,0.000731,0,0.000446,0.0508,0,0.00297 30,0,0.000117,0.000302,0.00283,0,0.301,0,0.0015266,0.000822,0.0288,0.001063,0.00222,0,0.000448,0.000908,0.000979,0.000734,0.000189,0,0.00157,0.00855,0.00325,0,0,0.00609,0.000064,0.00341,0,0.000709,0.000393,0.000111,0.00147,0.003,0,0,0,0,0.00218,0.000648,0.00871,0.0014,0,0.00192,0.000375,0.000277,0.000217,0.00064,0.0015,0.000266,0.0549,0,0.0306,0.0005,0.000897,0,0.00276,0.0502,0,0,0.00551,0.0509,0.00836,0.00338,0,0,0.00157,0.00257,0,0.00539,0.00258,0,0.000601,0.0037,0.0245,0.000563,0.000392,0.000552,0,0.00433,0.00181,0,0,0.0013,0.0513,0.00113,0.00289,0,0.00592,0.00042,0.000362,0,0 31,0,0,0,0.0501,0,0.354,0,0.000107,0.00056,0.0236,0,0.00195,0,0.000222,0,0,0,0,0,0.00039,0.00228,0.00186,0,0,0.00199,0,0.00169,0,0,0,0,0.000478,0.000401,0,0,0,0,0,0.000381,0.00477,0,0,0.000601,0,0,0,0,0,0,0.00765,0,0.0593,0.00093,0.000783,0,0.001654,0.00158,0,0,0.003025,0,0.00824,0.00244,0,0,0.00118,0.00107,0.000935,0.003,0,0,0,0.00167,0.00665,0,0,0,0,0.00209,0.00074,0,0,0.00164,0.00387,0.000659,0,0,0,0,0.00025,0,0 32,0,0,0,0.00676,0,0.237,0,0.000274,0.000929,0.043,0,0.00128,0,0.00029,0.000168,0.000599,0,0,0,0.000675,0.00444,0.00253,0,0,0.00401,0,0.00265,0,0,0,0,0.00153,0.00115,0,0,0,0.000206,0,0.000813,0.00762,0.00122,0,0.00152,0,0.00046,0,0,0.00125,0.000173,0.00612,0,0.0651,0,0.00224,0,0.000606,0.00529,0,0,0.00722,0.00018,0.00496,0.00425,0,0,0.00181,0.00219,0,0.00639,0.00517,0,0,0.0033,0.104,0,0.000853,0,0,0.00506,0.00069,0,0,0.00611,0.00994,0.00129,0.00092,0.00239,0.000906,0.000345,0,0,0 33,0,0,0,0.00564,0,0.239,0,0.000174,0.000588,0.0321,0,0.0014,0,0,0,0.000247,0,0,0,0.00113,0.00435,0.0015,0,0,0.00381,0,0.00217,0,0,0,0,0.000431,0.000707,0,0,0,0,0,0.0005,0.00646,0.00104,0,0.000834,0.000736,0,0,0,0,0.000117,0.00978,0,0.0333,0,0,0,0.000206,0.00329,0,0.00114,0.00303,0,0.00454,0.00307,0,0,0.00176,0.00167,0,0.00644,0.00257,0,0,0.00163,0.116,0,0.000308,0,0,0.00539,0.000727,0,0,0.00488,0.0506,0.00137,0.000872,0.00325,0.00256,0,0.00107,0.000295,0 34,0,0,0,0.00493,0,0.384,0,0.000221,0.000445,0.0582,0,0.000837,0,0,0,0,0,0,0,0,0.000938,0.000644,0,0,0.000695,0,0.000766,0,0,0,0,0,0,0,0,0,0,0.000691,0,0.00128,0,0,0,0,0,0,0,0,0,0.000638,0,0.0097,0,0,0,0.000495,0.000651,0,0,0.000926,0,0.00249,0,0,0,0,0.000751,0,0.000695,0,0,0,0,0.0997,0,0,0,0,0.000726,0,0.000766,0,0,0.00168,0.00015,0,0,0,0,0,0,0 35,0,0,0,0.00818,0,0.286,0,0.000202,0.000427,0.0568,0,0.00183,0,0,0,0,0,0,0,0,0.00223,0.0014,0,0,0.00183,0,0.00221,0,0,0,0,0,0.000418,0,0,0,0,0,0,0.00359,0.000571,0.00182,0.0004,0,0,0,0,0,0,0.00642,0,0.0382,0.000433,0.00139,0,0.000329,0.000436,0,0,0.00335,0,0.0531,0.00404,0,0,0,0.00122,0,0.00288,0,0,0,0.00197,0.00429,0,0,0,0,0.00708,0.000458,0,0,0.0019,0.00383,0.000668,0.0004,0,0,0,0,0,0 36,0,0,0,0.00117,0,0.141,0,0.000151,0.000289,0.00501,0,0.000397,0,0,0.0000811,0,0,0,0,0.00052,0.00423,0.000857,0,0,0.00353,0,0.00206,0,0,0,0,0.000198,0.00175,0,0,0,0,0,0,0.00462,0,0,0.00113,0.000325,0,0,0,0,0,0.0566,0,0.0476,0.000534,0,0,0.000507,0.00233,0,0,0.00375,0.00041,0.0755,0.00431,0,0,0.00124,0.00179,0,0.00618,0.00118,0,0.000205,0.00188,0.206,0.000504,0.00074,0,0,0.00365,0.000812,0,0,0.000763,0.00996,0.00113,0.00148,0,0.00185,0.00055,0.0064,0,0.00082 37,0.000266,0,0,0,0,0,0,0,0,0.00551,0,0,0,0,0,0,0,0,0.000821,0.00326,0.0274,0,0.00108,0,0,0,0,0,0,0,0.0005,0.00215,0,0,0.00345,0,0,0,0.00326,0.0267,0.00789,0.0097,0,0.0568,0,0,0,0.0016,0,0.000911,0,0,0.00175,0.00141,0,0,0.0201,0.0516,0,0.127,0,0.0217,0,0.00778,0,0.00581,0,0,0,0.00342,0,0.00397,0.03,0,0,0.00533,0,0,0.00143,0,0,0,0.00751,0,0,0,0,0,0.00294,0,0,0 38,0,0,0,0.00644,0,0.176,0,0.000172,0.000748,0.0759,0,0.00214,0,0.000474,0.000183,0.000378,0.000365,0,0,0.000887,0.00662,0.00241,0,0,0.00486,0,0.00356,0,0,0.000127,0,0.000828,0.00115,0,0,0,0,0,0.001,0.00983,0.00128,0,0.00169,0.00065,0,0.000265,0,0,0,0.0046,0,0.0438,0,0.000946,0,0.00264,0.0056,0,0,0.00994,0.00214,0.0088,0.00571,0,0,0.00208,0.0509,0.000342,0.00446,0.00355,0,0,0.00327,0.0253,0,0.00234,0,0,0.00337,0.000954,0,0,0.00702,0.059,0.00262,0.00117,0,0,0.00113,0.00163,0,0.000305 39,0,0,0,0.00994,0,0.265,0,0,0.000332,0.0321,0,0.00119,0,0,0,0,0,0,0,0,0.00189,0.000941,0,0,0.00168,0,0.000406,0,0,0,0,0.00057,0,0,0,0,0,0,0,0.0032,0.000428,0,0.000403,0.000237,0,0.000209,0,0,0,0.00194,0,0.0341,0,0,0,0,0,0.00077,0,0.003878,0,0.0919,0.00215,0,0,0,0.000813,0,0.00235,0.00115,0,0,0.00206,0.036,0.000332,0,0,0,0.00171,0,0,0,0.00494,0.00219,0.000405,0.000571,0,0,0,0.00209,0,0 40,0,0,0,0.00679,0,0.23,0,0,0.000376,0.0544,0,0.000633,0,0,0.000193,0.000227,0,0,0,0.0004,0.00371,0.00121,0,0,0.00351,0,0.00164,0,0,0,0,0.000887,0.000523,0,0,0,0,0,0.000558,0.00538,0.000757,0,0.00112,0.000532,0,0,0,0,0,0.0447,0,0.0201,0.00393,0,0.00632,0.000329,0.00139,0,0,0.0028,0,0.00265,0.00274,0,0,0,0.00144,0,0.00422,0,0,0,0.000919,0.0292,0.0042,0,0,0,0.05,0.000685,0,0,0.00246,0.00746,0.000854,0.00195,0,0.00086,0,0.00476,0,0 41,0,0,0,0.00693,0,0.267,0,0.00108,0.00044,0.027,0.00027,0.00307,0,0,0,0,0,0,0,0,0.00241,0.00302,0,0,0.00222,0,0.00138,0,0,0,0,0.000173,0,0,0.000351,0,0,0,0.00048,0.00311,0.000577,0,0.000406,0,0,0,0,0,0,0.0504,0,0.0246,0,0.00076,0,0.000852,0,0,0,0.00245,0.0013,0,0.00218,0,0,0,0.00082,0,0.00249,0.00262,0,0,0.00207,0.081,0.000357,0,0,0,0.00284,0,0,0,0.00178,0.00427,0.000877,0,0,0,0,0.000599,0,0.00201 42,0,0,0,0.0228,0,0.257,0,0,0.000132,0.0621,0,0.00101,0,0,0,0.000339,0,0,0,0,0.00376,0.00109,0,0,0.00218,0,0,0.000567,0,0,0,0.00141,0.000459,0,0,0,0,0,0.00055,0.0044,0.000964,0,0.000591,0.00132,0,0,0,0,0,0.00769,0,0.069,0.000146,0,0,0,0.00108,0,0,0.00765,0,0.00819,0,0,0.0583,0,0.00101,0,0.00423,0.00369,0,0,0.0563,0.0514,0,0.000903,0,0.000298,0,0.000407,0,0,0.00213,0.00236,0.000971,0,0.000561,0,0,0.00365,0,0.00053 ================================================ FILE: Demos/Mapped Network From Data (Biochemical and Structural)/edge.list.csv ================================================ "","source","target","value","type" "1",8,12,1,"KEGG" "2",11,12,1,"KEGG" "3",3,26,1,"KEGG" "4",8,26,1,"KEGG" "5",7,50,1,"KEGG" "6",12,50,1,"KEGG" "7",43,52,1,"KEGG" "8",30,56,1,"KEGG" "9",8,77,1,"KEGG" "10",26,82,1,"KEGG" "11",58,90,1,"KEGG" "12",60,90,1,"KEGG" "13",60,92,1,"KEGG" "213",3,5,0.829268292682927,"Tanimoto" "423",3,8,0.825,"Tanimoto" "424",5,8,0.702127659574468,"Tanimoto" "633",3,11,0.739130434782609,"Tanimoto" "634",5,11,0.775510204081633,"Tanimoto" "637",8,11,0.847826086956522,"Tanimoto" "703",3,12,0.794117647058823,"Tanimoto" "842",2,14,0.727272727272727,"Tanimoto" "923",14,15,0.794871794871795,"Tanimoto" "993",14,16,0.794871794871795,"Tanimoto" "994",15,16,0.897435897435897,"Tanimoto" "1053",3,17,0.75,"Tanimoto" "1054",5,17,0.790697674418605,"Tanimoto" "1123",3,18,0.82051282051282,"Tanimoto" "1127",8,18,0.853658536585366,"Tanimoto" "1130",11,18,0.729166666666667,"Tanimoto" "1131",12,18,0.72972972972973,"Tanimoto" "1191",1,19,0.884615384615385,"Tanimoto" "1261",1,20,0.785714285714286,"Tanimoto" "1274",15,20,0.72972972972973,"Tanimoto" "1275",16,20,0.72972972972973,"Tanimoto" "1278",19,20,0.892857142857143,"Tanimoto" "1331",1,21,0.785714285714286,"Tanimoto" "1344",15,21,0.72972972972973,"Tanimoto" "1345",16,21,0.72972972972973,"Tanimoto" "1348",19,21,0.892857142857143,"Tanimoto" "1349",20,21,1,"Tanimoto" "1483",14,25,0.846153846153846,"Tanimoto" "1484",15,25,0.948717948717949,"Tanimoto" "1485",16,25,0.948717948717949,"Tanimoto" "1543",3,26,0.825,"Tanimoto" "1544",5,26,0.702127659574468,"Tanimoto" "1547",8,26,0.733333333333333,"Tanimoto" "1557",18,26,0.727272727272727,"Tanimoto" "1693",14,28,0.846153846153846,"Tanimoto" "1694",15,28,0.948717948717949,"Tanimoto" "1695",16,28,0.948717948717949,"Tanimoto" "1702",25,28,1,"Tanimoto" "1762",13,29,0.702702702702703,"Tanimoto" "1832",13,29,0.702702702702703,"Tanimoto" "1846",29,29,1,"Tanimoto" "1966",7,32,0.808080808080808,"Tanimoto" "2184",15,36,0.755555555555556,"Tanimoto" "2185",16,36,0.755555555555556,"Tanimoto" "2192",25,36,0.8,"Tanimoto" "2195",28,36,0.8,"Tanimoto" "2313",3,38,0.702127659574468,"Tanimoto" "2314",5,38,0.74,"Tanimoto" "2381",1,39,0.709677419354839,"Tanimoto" "2398",19,39,0.806451612903226,"Tanimoto" "2399",20,39,0.9,"Tanimoto" "2400",21,39,0.9,"Tanimoto" "2413",37,39,0.75,"Tanimoto" "2451",1,40,0.709677419354839,"Tanimoto" "2468",19,40,0.806451612903226,"Tanimoto" "2469",20,40,0.9,"Tanimoto" "2470",21,40,0.9,"Tanimoto" "2483",37,40,0.75,"Tanimoto" "2485",39,40,1,"Tanimoto" "2551",35,42,0.716666666666667,"Tanimoto" "2553",37,42,0.826086956521739,"Tanimoto" "2689",32,45,0.776859504132231,"Tanimoto" "2731",1,46,0.709677419354839,"Tanimoto" "2749",20,46,0.727272727272727,"Tanimoto" "2750",21,46,0.727272727272727,"Tanimoto" "2821",22,47,0.774193548387097,"Tanimoto" "2834",38,47,0.811320754716981,"Tanimoto" "2968",30,50,0.732142857142857,"Tanimoto" "3013",3,51,0.717391304347826,"Tanimoto" "3014",5,51,0.755102040816326,"Tanimoto" "3044",38,51,0.895833333333333,"Tanimoto" "3051",47,51,0.9,"Tanimoto" "3102",25,52,0.7,"Tanimoto" "3105",28,52,0.7,"Tanimoto" "3106",29,52,0.74468085106383,"Tanimoto" "3107",29,52,0.74468085106383,"Tanimoto" "3156",7,53,0.724770642201835,"Tanimoto" "3179",32,53,0.858490566037736,"Tanimoto" "3189",45,53,0.75,"Tanimoto" "3248",30,56,0.834862385321101,"Tanimoto" "3249",32,56,0.845454545454545,"Tanimoto" "3266",53,56,0.752066115702479,"Tanimoto" "3323",37,57,0.8,"Tanimoto" "3325",39,57,0.731707317073171,"Tanimoto" "3326",40,57,0.731707317073171,"Tanimoto" "3327",42,57,0.931818181818182,"Tanimoto" "3393",37,58,0.883720930232558,"Tanimoto" "3395",39,58,0.731707317073171,"Tanimoto" "3396",40,58,0.731707317073171,"Tanimoto" "3397",42,58,0.847826086956522,"Tanimoto" "3408",57,58,0.906976744186046,"Tanimoto" "3461",35,59,0.728813559322034,"Tanimoto" "3463",37,59,0.844444444444444,"Tanimoto" "3467",42,59,0.977272727272727,"Tanimoto" "3478",57,59,0.953488372093023,"Tanimoto" "3479",58,59,0.866666666666667,"Tanimoto" "3531",35,60,0.728813559322034,"Tanimoto" "3533",37,60,0.844444444444444,"Tanimoto" "3537",42,60,0.977272727272727,"Tanimoto" "3548",57,60,0.953488372093023,"Tanimoto" "3549",58,60,0.866666666666667,"Tanimoto" "3550",59,60,1,"Tanimoto" "3692",61,61,1,"Tanimoto" "3728",19,62,0.735294117647059,"Tanimoto" "3729",20,62,0.818181818181818,"Tanimoto" "3730",21,62,0.818181818181818,"Tanimoto" "3745",39,62,0.909090909090909,"Tanimoto" "3746",40,62,0.909090909090909,"Tanimoto" "3747",42,62,0.75,"Tanimoto" "3758",57,62,0.804878048780488,"Tanimoto" "3759",58,62,0.804878048780488,"Tanimoto" "3760",59,62,0.767441860465116,"Tanimoto" "3761",60,62,0.767441860465116,"Tanimoto" "3798",19,64,0.735294117647059,"Tanimoto" "3799",20,64,0.818181818181818,"Tanimoto" "3800",21,64,0.818181818181818,"Tanimoto" "3815",39,64,0.909090909090909,"Tanimoto" "3816",40,64,0.909090909090909,"Tanimoto" "3817",42,64,0.75,"Tanimoto" "3828",57,64,0.804878048780488,"Tanimoto" "3829",58,64,0.804878048780488,"Tanimoto" "3830",59,64,0.767441860465116,"Tanimoto" "3831",60,64,0.767441860465116,"Tanimoto" "3834",62,64,1,"Tanimoto" "3868",19,66,0.735294117647059,"Tanimoto" "3869",20,66,0.818181818181818,"Tanimoto" "3870",21,66,0.818181818181818,"Tanimoto" "3885",39,66,0.909090909090909,"Tanimoto" "3886",40,66,0.909090909090909,"Tanimoto" "3887",42,66,0.75,"Tanimoto" "3898",57,66,0.804878048780488,"Tanimoto" "3899",58,66,0.804878048780488,"Tanimoto" "3900",59,66,0.767441860465116,"Tanimoto" "3901",60,66,0.767441860465116,"Tanimoto" "3904",62,66,1,"Tanimoto" "3905",64,66,1,"Tanimoto" "3934",15,68,0.755102040816326,"Tanimoto" "3935",16,68,0.755102040816326,"Tanimoto" "3942",25,68,0.795918367346939,"Tanimoto" "3945",28,68,0.795918367346939,"Tanimoto" "3951",35,68,0.830508474576271,"Tanimoto" "3952",36,68,0.784313725490196,"Tanimoto" "3965",52,68,0.727272727272727,"Tanimoto" "4004",15,69,0.755102040816326,"Tanimoto" "4005",16,69,0.755102040816326,"Tanimoto" "4012",25,69,0.795918367346939,"Tanimoto" "4015",28,69,0.795918367346939,"Tanimoto" "4021",35,69,0.830508474576271,"Tanimoto" "4022",36,69,0.784313725490196,"Tanimoto" "4035",52,69,0.727272727272727,"Tanimoto" "4047",68,69,1,"Tanimoto" "4103",50,71,0.723076923076923,"Tanimoto" "4402",77,78,0.844919786096257,"Tanimoto" "4453",50,80,0.792307692307692,"Tanimoto" "4457",56,80,0.77037037037037,"Tanimoto" "4483",3,82,0.75,"Tanimoto" "4484",5,82,0.714285714285714,"Tanimoto" "4497",18,82,0.702127659574468,"Tanimoto" "4503",26,82,0.863636363636364,"Tanimoto" "4653",37,85,1,"Tanimoto" "4655",39,85,0.75,"Tanimoto" "4656",40,85,0.75,"Tanimoto" "4657",42,85,0.826086956521739,"Tanimoto" "4668",57,85,0.8,"Tanimoto" "4669",58,85,0.883720930232558,"Tanimoto" "4670",59,85,0.844444444444444,"Tanimoto" "4671",60,85,0.844444444444444,"Tanimoto" "4793",37,90,0.8,"Tanimoto" "4797",42,90,0.843137254901961,"Tanimoto" "4808",57,90,0.82,"Tanimoto" "4809",58,90,0.82,"Tanimoto" "4810",59,90,0.86,"Tanimoto" "4811",60,90,0.86,"Tanimoto" "4827",85,90,0.8,"Tanimoto" "4863",37,92,0.730769230769231,"Tanimoto" "4867",42,92,0.843137254901961,"Tanimoto" "4878",57,92,0.82,"Tanimoto" "4879",58,92,0.75,"Tanimoto" "4880",59,92,0.86,"Tanimoto" "4881",60,92,0.86,"Tanimoto" "4897",85,92,0.730769230769231,"Tanimoto" "4899",90,92,0.923076923076923,"Tanimoto" "4831",1,92,0.431372549019608,"Tanimoto weak" "4832",2,92,0.321428571428571,"Tanimoto weak" "4840",11,92,0.333333333333333,"Tanimoto weak" "4842",13,92,0.327586206896552,"Tanimoto weak" "4843",14,92,0.406779661016949,"Tanimoto weak" "4844",15,92,0.474576271186441,"Tanimoto weak" "4845",16,92,0.474576271186441,"Tanimoto weak" "4848",19,92,0.490196078431373,"Tanimoto weak" "4849",20,92,0.54,"Tanimoto weak" "4850",21,92,0.54,"Tanimoto weak" "4852",25,92,0.459016393442623,"Tanimoto weak" "4855",28,92,0.459016393442623,"Tanimoto weak" "4857",29,92,0.34375,"Tanimoto weak" "4861",35,92,0.651515151515151,"Tanimoto weak" "4862",36,92,0.46031746031746,"Tanimoto weak" "4865",39,92,0.6,"Tanimoto weak" "4866",40,92,0.6,"Tanimoto weak" "4870",46,92,0.428571428571429,"Tanimoto weak" "4875",52,92,0.391304347826087,"Tanimoto weak" "4884",62,92,0.66,"Tanimoto weak" "4885",64,92,0.66,"Tanimoto weak" "4886",66,92,0.66,"Tanimoto weak" "4887",68,92,0.523076923076923,"Tanimoto weak" "4888",69,92,0.523076923076923,"Tanimoto weak" "4890",72,92,0.530120481927711,"Tanimoto weak" "4896",84,92,0.333333333333333,"Tanimoto weak" "4898",87,92,0.310126582278481,"Tanimoto weak" "4828",87,90,0.30188679245283,"Tanimoto weak" ================================================ FILE: Demos/Mapped Network From Data (Biochemical and Structural)/network code.r ================================================ setwd("C:\\Users\\D\\Dropbox\\Software\\TeachingDemos\\Demos\\Mapped Network From Data (Biochemical and Structural)") setwd("C:\\Users\\Node\\Dropbox\\Software\\TeachingDemos\\Demos\\Mapped Network From Data (Biochemical and Structural)") setwd("C:\\Users\\dgrapov\\Dropbox\\Software\\TeachingDemos\\Demos\\Mapped Network From Data (Biochemical and Structural)") source("http://pastebin.com/raw.php?i=1Bs7G5ds") #source devium source("http://pastebin.com/raw.php?i=UyDBTA57") # #save() # load objects tmp.data<-read.csv("data.csv",row.names=1) var.meta<-read.csv("variable info.csv") sample.meta<-read.csv("sample info.csv") # Calculate Network Connections #--------------------------------- #identify required Pubchem CIDs and KEGG IDs CIDS<-fixln(var.meta$PubChem) KEGG<-fixlc(var.meta$KEGG) #get KEGG RPAIRS # load reaction DB and return all reactions of type main reaction.DB<-get.KEGG.pairs(type="main") kegg.edges<-get.Reaction.pairs(KEGG,reaction.DB,index.translation.DB=NULL,parallel=FALSE,translate=FALSE) #create shared index to allow merging with other edge identifiers index<-KEGG edge.names<-data.frame(index, network.id = c(1:length(index))) kegg.edges<-data.frame(make.edge.list.index(edge.names,kegg.edges)) #get structural similarity edges based on Tanimoto >0.7 tanimoto.edges<-CID.to.tanimoto(cids=CIDS, cut.off = 0, parallel=FALSE) #return all possible connections #create shared index index<-CIDS edge.names<-data.frame(index, network.id = c(1:length(index))) tmp<-make.edge.list.index(edge.names,tanimoto.edges) tanimoto.edges[,1:2]<-tmp #prepare tanimoto edges for merge with KEGG tanimoto.edges$type<-"Tanimoto" #merge the biochemical and structural similarity edge lists kegg.edges$value<-1 # give arbitrary weight, here the max tanimoto can take kegg.edges$type<-"KEGG" # set type to identify between KEGG and tanimoto final.edge.list<-rbind(kegg.edges,tanimoto.edges) #note duplicated edges maybe prioritized diffrently in various orders #write.csv(final.edge.list,file="edge.list.csv") # uncomment to save the file #render network for preview { edge.list<-clean.edgeList(data=final.edge.list) tmp.edge.list<-edge.list[fixln(edge.list$value)>=0.7,] ggplot2.network(tmp.edge.list,edge.color.var="type", bezier=FALSE,node.size=3, node.names=fixlc(var.meta$Name2),node.label.size = 3) #removing unconnected nodes and getting names right tmp.edge.list$source<-paste(" ",tmp.edge.list$source,sep="") tmp.edge.list$target<-paste(" ",tmp.edge.list$target,sep="") ggplot2.network(tmp.edge.list,edge.color.var="type", bezier=FALSE,node.size=3, node.names=node.names,node.label.size = 3) #create node attributes id<-unique(unlist(tmp.edge.list[,1:2])) node.names<-data.frame(fixlc(var.meta$Name2)[fixln(gsub(" ","",id))])#[order(id,decreasing=TRUE)] fct<-rep(c(1:2),length.out=nrow(node.names)) len<-length(unique(fct)) node.data<-data.frame(name=node.names[,],color=rainbow(len)[fct],size=seq(1,6,by=.5)[fct]) rownames(node.data)<-unique(unlist(tmp.edge.list[,1:2])) ggplot2.network(edge.list=tmp.edge.list,edge.color.var="type", bezier=FALSE,node.size=3, node.data=node.data,node.names=node.names, max.edge.thickness = 1,node.label.size = 3) #create function to remove duplicated edges and self edges # from an edgelist with diffrent types controling heirarchy of existence edge.list<-data.frame(data.frame(source=c(3,2,3,4),target=c(3,2,1,3))) edge.list$type<-c("a","b","b","b") edge.list$extra<-c(1:4) clean.edgeList(data=edge.list) clean.edgeList<-function(source="source",target="target",type="type", data=edge.list){ library(igraph) #remove self edges else if all self passed will cause an error el<-data[,c(source,target)] self<-el[,1]==el[,2] el<-el[!self,] tmp.data<-as.data.frame(as.matrix(data)[!self,]) lel<-split(el,tmp.data$type) el.res<-do.call("rbind",lapply(1:length(lel),function(i){ nodes<-matrix(sort(unique(matrix(as.matrix(lel[[i]]),,1))),,1) g<-graph.data.frame(lel[[i]],directed=FALSE,vertices=nodes) g.adj<-get.adjacency(g,sparse=FALSE,type="upper") g.adj[g.adj>0]<-1 adj<-graph.adjacency(g.adj,mode="upper",diag=FALSE,add.rownames="code") get.edgelist(adj) })) ids<-unique(join.columns(el.res)) tmp<-data.frame(el,tmp.data[,!colnames(tmp.data)%in%c(source,target)]) rownames(tmp)<-make.unique(join.columns(tmp[,1:2])) flip<-!ids%in%rownames(tmp) ids[flip]<-unique(join.columns(el.res[,2:1])) return(tmp[ids,]) } #over lay paths and nodes as separate images? .theme<- theme( axis.line = element_blank(), axis.ticks = element_blank(), axis.title.x = element_blank(), panel.background = element_blank(), plot.background = element_blank(), panel.grid = element_blank(), axis.text.x = element_blank(), axis.text.y = element_blank(), axis.title.x = element_blank(), axis.title.y = element_blank(), legend.key = element_blank() ) vis<-data.frame(x=1:2,y=3:4) p<-ggplot(vis, aes(x=x,y=y)) p+geom_line()+.theme png(file ="layer1.png", pointsize=1,width=600,height=600, bg = "transparent") p+geom_line()+.theme dev.off() png(file ="layer2.png", pointsize=1,width=600,height=600, bg = "transparent") p+geom_point(color="red",size=2)+.theme dev.off() library(png) i1 <- readPNG("layer1.png", native=FALSE) i2 <- readPNG("layer2.png", native=FALSE) ghostize <- function(r, alpha=0.5) matrix(adjustcolor(rgb(r[,,1],r[,,2],r[,,3],r[,,4]), alpha.f=alpha), nrow=dim(r)[1]) grid.newpage() grid.rect(gp=gpar(fill="white")) grid.raster(i1) grid.raster(i2) library(png) img <- readPNG("layer1.png") r = as.raster(img[,,1:3]) r[img[,,4] == 0] = "white" plot(1:2,type="n") rasterImage(r,1,1,2,2) N <- 1000 # Warning: slow d <- data.frame(x1=rnorm(N), x2=rnorm(N, 0.8, 0.9), y=rnorm(N, 0.8, 0.2), z=rnorm(N, 0.2, 0.4)) v <- with(d, dataViewport(c(x1,x2),c(y, z))) png("layer1.png", bg="transparent") with(d, grid.points(x1,y, vp=v,default="native",pch=".",gp=gpar(col="blue"))) dev.off() png("layer2.png", bg="transparent") with(d, grid.points(x2,z, vp=v,default="native",pch=".",gp=gpar(col="red"))) dev.off() library(png) i1 <- readPNG("layer1.png", native=FALSE) i2 <- readPNG("layer2.png", native=FALSE) ghostize <- function(r, alpha=0.5) matrix(adjustcolor(rgb(r[,,1],r[,,2],r[,,3],r[,,4]), alpha.f=alpha), nrow=dim(r)[1]) grid.newpage() grid.rect(gp=gpar(fill="white")) grid.raster(ghostize(i1)) grid.raster(ghostize(i2)) #function to plot network in ggplot 2 #ggplot based network drawing fxn ggplot2.network<-function(edge.list, edge.color.var = NULL, edge.color = NULL, directed = FALSE, node.data=NULL, node.color = NULL, node.names=NULL, show.names = TRUE, node.shape=15, bezier = FALSE, node.size = 7,node.label.size = 5, max.edge.thickness = 2, color.scale=NULL,fill.scale=NULL, group.bounds=NULL){ # edge list = 2 column data.frame representing source and target. # Columns over 2 will be sorted with edgelist and can be segment mapped to color transparency and width # edge.color.var = name of variable in edge list to use to color # edge.color = color for each level of object edge.color.var # directed = logical, if FALSE edge will be transposed and duplicated making undirected # node.color = colors for nodes, need to take into account node name ordering # show.names = can be supplied names for nodes, TRUE = network index, FALSE = nothing # node names should be a 2 column matrix with edge IDs and mapped names #should have a global node attributes (names, color, size, etc) object form which colnames are used for various mappings library(network) # as.network library(sna) # layouts library(ggplot2) library(Hmisc) # bezier edges # Function to generate paths between each connected node (very slow when transparent!) # adapted from : https://gist.github.com/dsparks/4331058 edgeMaker <- function(whichRow, len = 100, curved = TRUE){ fromC <- layoutCoordinates[adjacencyList[whichRow, 1], ] # Origin toC <- layoutCoordinates[adjacencyList[whichRow, 2], ] # Terminus # Add curve: graphCenter <- colMeans(layoutCoordinates) # Center of the overall graph bezierMid <- c(fromC[1], toC[2]) # A midpoint, for bended edges distance1 <- sum((graphCenter - bezierMid)^2) if(distance1 < sum((graphCenter - c(toC[1], fromC[2]))^2)){ bezierMid <- c(toC[1], fromC[2]) } # To select the best Bezier midpoint bezierMid <- (fromC + toC + bezierMid) / 3 # Moderate the Bezier midpoint if(curved == FALSE){bezierMid <- (fromC + toC) / 2} # Remove the curve edge <- data.frame(bezier(c(fromC[1], bezierMid[1], toC[1]), # Generate c(fromC[2], bezierMid[2], toC[2]), # X & y evaluation = len)) # Bezier path coordinates edge$Sequence <- 1:len # For size and colour weighting in plot edge$Group <- paste(adjacencyList[whichRow, 1:2], collapse = ">") if(ncol(adjacencyList)>2){ tmp<-data.frame(matrix(as.matrix(adjacencyList[whichRow, -c(1,2),drop=FALSE]),nrow = nrow(edge), ncol=ncol(adjacencyList)-2, byrow=TRUE)) colnames(tmp)<-colnames(adjacencyList)[-c(1:2)] edge$extra<-tmp edge<-do.call("cbind",edge) colnames(edge)<-gsub("extra.","",colnames(edge)) } return(edge) } #straight edges edgeMaker2<-function(whichRow){ fromC <- layoutCoordinates[adjacencyList[whichRow, 1], ] # Origin toC <- layoutCoordinates[adjacencyList[whichRow, 2], ] # Terminus edge <- data.frame(c(fromC[1], toC[1]), c(fromC[2] ,toC[2])) # Generate # X & ) # Bezier path coordinates edge$Sequence <- 1 # For size and colour weighting in plot edge$Group <- paste(adjacencyList[whichRow, 1:2], collapse = ">") #get other info if supplied with edge list if(ncol(adjacencyList)>2){ tmp<-data.frame(matrix(as.matrix(adjacencyList[whichRow, -c(1,2),drop=FALSE]),nrow = nrow(edge), ncol=ncol(adjacencyList)-2, byrow=TRUE)) colnames(tmp)<-colnames(adjacencyList)[-c(1:2)] edge$extra<-tmp edge<-do.call("cbind",edge) colnames(edge)<-gsub("extra.","",colnames(edge)) } colnames(edge)[1:2]<-c("x","y") return(edge) } # adding transposed source target edges to make undirected bezier curves if (bezier == TRUE) { if(all(!directed)) { is.rev<-rep(TRUE, nrow(edge.list)) } else { is.rev<-directed==TRUE } rev.edge.list<-data.frame(rbind(as.matrix(edge.list[,1:2]),as.matrix(edge.list[is.rev,2:1]))) # need matrix else no reordering of columns? } else{ rev.edge.list<-edge.list[,1:2,drop=FALSE] } #extra info (separate now, later recombine) info<-edge.list[,-c(1:2)] #getting layout and making sure edge list ids are in the same order g<-as.network(rev.edge.list[,1:2],matrix.type = "edgelist") # #layout node.layout<-gplot.layout.fruchtermanreingold(g[,], layout.par = NULL) n.edge.list<-as.matrix.network.edgelist(g) dimnames(node.layout)<-list(rownames(g[,]),c("x","y")) #preparing for edge path layoutCoordinates<-node.layout adjacencyList<-data.frame(n.edge.list,info) if (bezier == TRUE) { allEdges <- lapply(1:nrow(adjacencyList), edgeMaker, len = 500, curved = TRUE) allEdges <- do.call(rbind, allEdges) # a fine-grained path ^, with bend ^ } else { #straight edges using same controls(faster) allEdges <- lapply(1:nrow(adjacencyList), edgeMaker2) allEdges <- do.call(rbind, allEdges) } allEdges$neg.Sequence<- - allEdges$Sequence #Edge Attributes #------------------- #set default plotting variables # Edge colors edge.guide = TRUE if(is.null(edge.color.var)){edge.list$edge.color.var<-1;edge.color.var<-"edge.color.var";edge.guide = FALSE} if(is.null(edge.color)){ edge.color<-rainbow(nlevels(as.factor(with (edge.list, get(edge.color.var))))) } #Node Attributes #------------------- node.obj<-tryCatch(data.frame(layoutCoordinates,node.data[rownames(node.layout),]), error=function(e){data.frame(layoutCoordinates)}) #set defaults default<-factor(rep(1,nrow(node.obj))) # could match input column to those below here attribute<-c("size","color","shape") for(i in 1:length(attribute)){if(is.null(node.obj[[attribute[i]]])){node.obj[[attribute[i]]]<-default} }#else {node.obj[[attribute[i]]]<-factor(node.obj[[attribute[i]]])}} #default input #color if(is.null(node.color)){ node.color<-rainbow(length(unique(node.obj$color))) col.scale<-scale_color_manual(values=node.color) } #size if(is.null(node.size)){ node.size<-seq(3,7,length.out=length(unique(node.obj$size))) } #shape if(is.null(node.shape)){ node.shape<-rep(c(15:18),length.out=length(unique(node.obj$shape))) } node.points<-geom_point(data = node.obj, aes(x = x, y = y, color=color, size=size),shape=15,show_guide = TRUE) #,shape=shape # # testing # zp1+node.points # zp1 <- ggplot() # Pretty simple plot code # # bezier edges # zp1 <- zp1 + geom_path(data=allEdges,aes_string(x = "x", y = "y", group = "Group", # Edges with gradient # colour = edge.color.var, size = "neg.Sequence")) # and tap # zp1<-zp1+geom_point(data = node.obj, aes(x = x, y = y, color=color,size=size, shape=shape,fill=color),shape=21) # col.scale#,size=node.size,show_guide = FALSE) #labels if(is.null(node.obj$name)){ node.obj$name<-rownames(node.obj) } if(show.names==FALSE){node.obj$name<-rep("",nrow(node.layout))} #nothing node.labels<-geom_text(data = node.obj, aes(x = x, y = y-.2, label = name), size = node.label.size) # node names polygons<-NULL # #add grouping vis # #Hoettellings T2 ellipse # polygons<-NULL # if(group.bounds=="ellipse"){ # ell<-get.ellipse.coords(cbind(x=node.obj$x,y=node.obj$y), group=node.obj$group)# group visualization via # polygons<-if(is.null(color)){ # geom_polygon(data=data.frame(ell$coords),aes(x=x,y=y), fill="gray", color="gray",linetype=2,alpha=g.alpha, show_guide = FALSE) # } else { # geom_polygon(data=data.frame(ell$coords),aes(x=x,y=y, fill=group),linetype=2,alpha=g.alpha, show_guide = FALSE) # } # } # # # if(group.bounds=="polygon"){ # ell<-get.polygon.coords(data.frame(tmp.obj),tmp$color)# group visualization via # polygons<-if(is.null(color)){ # geom_polygon(data=data.frame(ell),aes(x=x,y=y), fill="gray", color="gray",linetype=2,alpha=g.alpha, show_guide = FALSE) # } else { # geom_polygon(data=data.frame(ell),aes(x=x,y=y, fill=group),linetype=2,alpha=g.alpha, show_guide = FALSE) # } # } #set up for plotting #theme new_theme_empty <- theme_bw() new_theme_empty$line <- element_blank() new_theme_empty$rect <- element_blank() new_theme_empty$strip.text <- element_blank() new_theme_empty$axis.text <- element_blank() new_theme_empty$plot.title <- element_blank() new_theme_empty$axis.title <- element_blank() new_theme_empty$plot.margin <- structure(c(0, 0, -1, -1), unit = "lines", valid.unit = 3L, class = "unit") new_theme_empty$legend.text <-element_text( size = 20) new_theme_empty$legend.title <-element_text(size = 20 ) # # node names (set above) # if(length(show.names) == attr(n.edge.list,"vnames")) { node.names <- show.names} # if (show.names) { node.names<-attr(n.edge.list,"vnames") } # if(!show.names){node.names<-rep("",nrow(node.layout))} #make plot zp1 <- ggplot() # Pretty simple plot code #area zp1<-zp1 + polygons #edges zp1 <- zp1 + geom_path(data=allEdges,aes_string(x = "x", y = "y", group = "Group", # Edges with gradient colour = edge.color.var),size=max.edge.thickness) # and taper # Customize taper #nodes zp1 <- zp1 + node.points + node.labels # node.obj<-data.frame(layoutCoordinates, color = as.factor(node.color), shape=as.factor(node.shape)) # zp1 <- zp1 + geom_point(data = node.obj, aes(x = x, y = y, fill=color, shape=shape), size = node.size, colour = "black", show_guide = node.guide)# Add # zp1<-zp1 + scale_fill_manual(values=fixlc(node.obj$color)) + scale_shape_manual(values =fixln(node.obj$shape)) # zp1<-zp1 + geom_text(data = data.frame(layoutCoordinates, label = node.names), aes(x = x, y = y-.2, label = label), size = node.label.size) # node names zp1 <- zp1 + scale_colour_manual(values = c(node.color,edge.color) ) # zp1 <- zp1 + scale_size(range = c(1/100, max.edge.thickness), guide = "none") #edge thickness zp1 <-zp1 + guides(color = guide_legend(override.aes = list (size = 3))) + labs(color='Edge Type') # Customize gradient zp1 <- zp1 + new_theme_empty # Clean up plot print(zp1) } } #create Node attribute mappings #--------------------------------- # color = direction or fold change or O-PLS-DA absolute loading # size = log fold change or absolute value of O-PLS-DA loading on LV 1 # border = O-PLS-DA VIP >=1 # shape = chemical class # calculate summary statistics (fold change and p-value) for main hypothesis in study { # set main factor to sample.meta$group test.data<-log(tmp.data+1)#test shifted log transformed data p.values<-multi.t.test(data=test.data, factor=sample.meta$group,paired=FALSE,progress=TRUE) fold.change<-calc.FC(data=tmp.data,factor=sample.meta$group,denom=levels(sample.meta$group)[2],sig.figs=1,log=FALSE) #fold change will need scaling and has some problems with 0 values fold.change[fold.change=="Inf"]<-0 #create summary table stats<-data.frame(name=fixlc(var.meta$Name2),p.values,fold.change) write.csv(stats,file="statistics.csv") } # do PCA { pca.data<-tmp.data #set PCA args pca.inputs<-list() pca.inputs$pca.algorithm<-"svd" pca.inputs$pca.components<-4 pca.inputs$pca.center<-TRUE pca.inputs$pca.scaling<-"uv" pca.inputs$pca.data<-pca.data pca.inputs$pca.cv<-"q2" #calculate model res<-devium.pca.calculate(pca.inputs,return="list",plot=FALSE) # need to recreate model in another format to plot #plot scores by type results<-"scores"#"biplot"#"scores","loadings","biplot")"screeplot" color<-data.frame(group=sample.meta$group) xaxis<-1 yaxis=2 group.bounds="ellipse" plot.PCA(pca=res,results=results,yaxis=yaxis,xaxis=xaxis,size=4,color=color, label=T, legend.name = NULL,font.size=1.75,group.bounds,alpha=.75) # samples 37 and 13 maybe outliers } #O-PLS-DA { #calculate preliminary model comp<-3 ocomp<-2 pls.y<-data.frame(group=as.numeric(sample.meta$group)) scaled.data<-data.frame(scale(pca.data,center=TRUE,scale=TRUE)) mods1<-make.OSC.PLS.model(pls.y=pls.y,pls.data=scaled.data,comp=comp,OSC.comp=ocomp,validation = "LOO",method="oscorespls",cv.scale=T) plot.OSC.results(mods1,plot="scores",groups=color) final<-get.OSC.model(obj=mods1,OSC.comp=ocomp) #collect results for mapping node.obj<-data.frame() # plot scores plot.PLS(obj=final,results="scores",color=color,group.bounds="ellipse") #carry out feature selection #feature selection obj<-final type<-"quantile"#"number" top<-0.9 p.value=0.05 FDR=FALSE separate=FALSE .scores<-obj$scores[,] .loadings<-obj$loadings[,] selected.features<-PLS.feature.select(pls.data=scaled.data,pls.scores=.scores[,1],pls.loadings=.loadings[,1],pls.weight=.loadings[,1], p.value=p.value, FDR=FDR,cut.type=type,top=top,separate=separate,type="spearman",make.plot=FALSE) selected.features<-selected.features[,c(1,3,7)] selected.features$VIP<-as.matrix(obj$VIP[,1,drop=F]) } #map node attributes { tmp<-list() tmp$ID<-1:length(selected.features$VIP) tmp$color<-ifelse(fold.change>=1,"up","down") tmp$size<-selected.features$VIP tmp$selected<-selected.features$VIP>=1 tmp$name<-fixlc(var.meta$Name2) node.attributes<-data.frame(do.call("cbind",tmp)) colnames(node.attributes)<-names(tmp) #get db info for analytes DB<-IDEOMgetR() info<-enrichR.IDEOM(id=CIDS, from="PubChem CID",IDEOM.DB=DB) tmp<-fixlc(info$Map) tmp[is.na(tmp)]<-"other" node.attributes$class<-tmp write.csv(node.attributes,file="node.attributes.csv") } ================================================ FILE: Demos/Mapped Network From Data (Biochemical and Structural)/node.attributes.csv ================================================ ID,color,size,selected,name,class 1,down,1.299818383,TRUE,Propylene glycol,Lipid Metabolism 2,up,1.733627798,TRUE,Glycolic acid,Nucleotide Metabolism 3,up,0.906013278,FALSE,L-Alanine,other 4,down,0.252883372,FALSE,,other 5,down,0.788061117,FALSE,L-Valine,Amino Acid Metabolism 6,down,1.234377918,TRUE,Urea,Amino Acid Metabolism 7,down,0.790245582,FALSE,Benzoic acid,other 8,up,1.098129963,TRUE,L-Serine,Amino Acid Metabolism 9,down,0.634914989,FALSE,Ethanolamine,Amino Acid Metabolism 10,up,0.659531582,FALSE,Phosphoric acid,Energy Metabolism 11,up,0.962800591,FALSE,L-Threonine,Amino Acid Metabolism 12,up,1.266322545,TRUE,Glycine,Xenobiotics Biodegradation and Metabolism 13,down,0.790245582,FALSE,Succinic acid,other 14,up,0.521221946,FALSE,Glyceric acid,Carbohydrate Metabolism 15,up,0.546569114,FALSE,"2,4-Dihydroxybutanoic acid",other 16,up,1.93931343,TRUE,"(S)-3,4-Dihydroxybutyric acid",other 17,up,0.908627549,FALSE,3-Aminoisobutanoic acid,other 18,up,1.101488745,TRUE,Aminomalonic acid, 19,down,1.486559547,TRUE,"(2R*,3R*)-1,2,3-Butanetriol",other 20,up,1.286588241,TRUE,D-Threitol,other 21,up,0.990993538,FALSE,Erythritol,other 22,up,1.375844298,TRUE,Pyroglutamic acid,Amino Acid Metabolism 23,up,0.872181606,FALSE,,other 24,down,0.645797177,FALSE,,other 25,up,1.247175218,TRUE,Erythronic acid,other 26,up,0.563686189,FALSE,L-Cysteine,Amino Acid Metabolism 27,down,0.538705285,FALSE,Creatinine,Amino Acid Metabolism 28,up,0.457205694,FALSE,Threonic acid,other 29,up,0.713549876,FALSE,Oxoglutaric acid,Carbohydrate Metabolism 30,up,1.345520899,TRUE,L-Phenylalanine,other 31,up,0.908828556,FALSE,,other 32,up,1.45892177,TRUE,p-Hydroxyphenylacetic acid,Amino Acid Metabolism 33,up,1.68358479,TRUE,,other 34,up,1.037666091,TRUE,Taurine,Lipid Metabolism 35,up,1.19678404,TRUE,Pectin, 36,down,0.842423667,FALSE,Arabinose,Amino Acid Metabolism 37,up,1.087402931,TRUE,D-Ribose,Carbohydrate Metabolism 38,up,0.961460011,FALSE,L-Lysine,Carbohydrate Metabolism 39,up,1.148532611,TRUE,D-Xylitol,Amino Acid Metabolism 40,up,1.296995751,TRUE,D-Arabitol,other 41,up,1.116369463,TRUE,Ribitol,other 42,up,0.75269885,FALSE,L-Fucose,Carbohydrate Metabolism 43,up,1.851909456,TRUE,cis-Aconitic acid,other 44,down,0.981482424,FALSE,D-Ribose,Carbohydrate Metabolism 45,up,2.048365046,TRUE,Homovanillic acid,Amino Acid Metabolism 46,down,0.440987815,FALSE,Glycerol 3-phosphate,Lipid Metabolism 47,up,0.721778091,FALSE,L-Glutamine,Amino Acid Metabolism 48,down,0.227845964,FALSE,,other 49,up,1.033067548,TRUE,Hypoxanthine,Carbohydrate Metabolism 50,up,0.76290555,FALSE,Hippuric acid,Amino Acid Metabolism 51,down,0.842423667,FALSE,Ornithine,Amino Acid Metabolism 52,up,0.661796848,FALSE,Citric acid,other 53,up,1.027152014,TRUE,3-(3-Hydroxyphenyl)-3-hydroxypropanoic acid,other 54,down,0.864456119,FALSE,,other 55,down,0.683583279,FALSE,Quinic acid,other 56,up,1.299131713,TRUE,L-Tyrosine,other 57,up,0.380070114,FALSE,L-Sorbose, 58,down,1.206093428,TRUE,D-Fructose,Carbohydrate Metabolism 59,up,0.900822861,FALSE,D-Galactose,Xenobiotics Drugs etc 60,up,0.793094738,FALSE,D-Glucose,other 61,up,0.949160549,FALSE,L-Histidine,other 62,up,0.608357131,FALSE,Mannitol,Amino Acid Metabolism 63,up,1.406634949,TRUE,,other 64,down,0.474463577,FALSE,Sorbitol,Carbohydrate Metabolism 65,up,0.878501982,FALSE,,other 66,down,0.688496547,FALSE,Galactitol,Carbohydrate Metabolism 67,down,0.620119883,FALSE,,other 68,up,0.767222179,FALSE,Gulonic acid,other 69,up,0.353143318,FALSE,Gluconic acid,Carbohydrate Metabolism 70,up,0.58165814,FALSE,,other 71,up,0.469785178,FALSE,Kynurenic acid,Amino Acid Metabolism 72,down,1.156206268,TRUE,Beta-N-Acetylglucosamine, 73,down,0.480199886,FALSE,,other 74,down,0.583574412,FALSE,Uric acid,Nucleotide Metabolism 75,up,0.798486399,FALSE,,other 76,down,0.596285854,FALSE,,other 77,up,1.527657596,TRUE,L-Tryptophan,Carbohydrate Metabolism 78,up,0.998177161,FALSE,5-Hydroxyindoleacetic acid,Amino Acid Metabolism 79,down,0.663749299,FALSE,,other 80,up,1.134615214,TRUE,4-Hydroxyhippuric acid,other 81,down,0.705675506,FALSE,,other 82,down,0.847647221,FALSE,L-Cystine,other 83,down,0.812004248,FALSE,,other 84,up,0.427448727,FALSE,Pseudouridine,Nucleotide Metabolism 85,up,1.540686424,TRUE,Arabinofuranose, 86,up,0.905064018,FALSE,,other 87,down,0.347224178,FALSE,Acetaminophen glucuronide,Carbohydrate Metabolism 88,up,0.817887212,FALSE,Hydroxyproline dipeptide 4-TMS,other 89,up,1.54246661,TRUE,Xylobiose,other 90,up,0.62092751,FALSE,Sucrose,Amino Acid Metabolism 91,up,0.979388644,FALSE,,other 92,up,0.54023557,FALSE,Alpha-Lactose, ================================================ FILE: Demos/Mapped Network From Data (Biochemical and Structural)/sample info.csv ================================================ ID,sample,class,group 1,A019,2,A 2,A021,,A 3,A023,3,A 4,A029,2,A 5,A031,3,A 6,A037,1,A 7,A040,2,A 8,A042,3,A 9,A051,,A 10,A053,2,A 11,A054,1,A 12,A055,2,A 13,A057,3,A 14,A061,3,A 15,A063,3,A 16,A065,1,A 17,A067,,A 18,A069,3,A 19,A071,3,A 20,A073,2,A 21,A077,3,A 22,C020,0,C 23,C022,0,C 24,C024,0,C 25,C030,0,C 26,C032,0,C 27,C038,0,C 28,C041,0,C 29,C043,0,C 30,C052,0,C 31,C054,0,C 32,C056,0,C 33,C058,0,C 34,C060,0,C 35,C062,0,C 36,C064,0,C 37,C066,0,C 38,C068,0,C 39,C070,0,C 40,C072,0,C 41,C074,0,C 42,C078,0,C ================================================ FILE: Demos/Mapped Network From Data (Biochemical and Structural)/statistics.csv ================================================ "","name","t.test_p.value","t.test_adjusted.p.value","t.test_q.value","FC.A.C" "X1","Propylene glycol",0.164881412830391,0.619253978472208,0.28788774182491,0 "X2","Glycolic acid",0.0198760115978473,0.619253978472208,0.28788774182491,3.6 "X3","L-Alanine",0.253768829006344,0.619253978472208,0.28788774182491,2.8 "X4","",0.805517644591721,0.871854391793393,0.405320273715117,0.9 "X5","L-Valine",0.329256577171709,0.619253978472208,0.28788774182491,0 "X6","Urea",0.263544795092355,0.619253978472208,0.28788774182491,0.9 "X7","Benzoic acid",0.329256577171709,0.619253978472208,0.28788774182491,0 "X8","L-Serine",0.205624719892993,0.619253978472208,0.28788774182491,1.7 "X9","Ethanolamine",0.585937463130809,0.792738920706389,0.368539929774745,0.9 "X10","Phosphoric acid",0.939932342558587,0.96206209129645,0.447257333157555,1 "X11","L-Threonine",0.231750325772512,0.619253978472208,0.28788774182491,2.6 "X12","Glycine",0.123976364637514,0.619253978472208,0.28788774182491,1.6 "X13","Succinic acid",0.329256577171709,0.619253978472208,0.28788774182491,0 "X14","Glyceric acid",0.728385365088196,0.84824624795081,0.394344978511868,1.1 "X15","2,4-Dihydroxybutanoic acid",0.768515517290104,0.849001881328481,0.394696268280376,1.1 "X16","(S)-3,4-Dihydroxybutyric acid",0.0497484671907041,0.619253978472208,0.28788774182491,1.9 "X17","3-Aminoisobutanoic acid",0.198594350833866,0.619253978472208,0.28788774182491,6.9 "X18","Aminomalonic acid",0.139796023200647,0.619253978472208,0.28788774182491,6.3 "X19","(2R*,3R*)-1,2,3-Butanetriol",0.329256577171709,0.619253978472208,0.28788774182491,0 "X20","D-Threitol",0.148967215292983,0.619253978472208,0.28788774182491,1.7 "X21","Erythritol",0.348730848839896,0.619253978472208,0.28788774182491,1.6 "X22","Pyroglutamic acid",0.20655427500011,0.619253978472208,0.28788774182491,1.2 "X23","",0.760260252307975,0.849001881328481,0.394696268280376,1.3 "X24","",0.438150043451881,0.639838158691636,0.297457213102767,0.2 "X25","Erythronic acid",0.189274253243826,0.619253978472208,0.28788774182491,1.9 "X26","L-Cysteine",0.60227083056255,0.793505619284379,0.368896363693543,2.3 "X27","Creatinine",0.76296861396891,0.849001881328481,0.394696268280376,0.9 "X28","Threonic acid",0.986866731047075,0.986866731047075,0.458788872675808,1 "X29","Oxoglutaric acid",0.463394223837754,0.666129196766771,0.309679770962398,1.7 "X30","L-Phenylalanine",0.093398637603731,0.619253978472208,0.28788774182491,2.2 "X31","",0.254758495997756,0.619253978472208,0.28788774182491,2.4 "X32","p-Hydroxyphenylacetic acid",0.102648972831718,0.619253978472208,0.28788774182491,2.1 "X33","",0.103344825685762,0.619253978472208,0.28788774182491,1.7 "X34","Taurine",0.165062457353588,0.619253978472208,0.28788774182491,3.9 "X35","Pectin",0.294850444096769,0.619253978472208,0.28788774182491,2.2 "X36","Arabinose",0.329256577171709,0.619253978472208,0.28788774182491,0 "X37","D-Ribose",0.158874692105942,0.619253978472208,0.28788774182491,5.7 "X38","L-Lysine",0.251540076711914,0.619253978472208,0.28788774182491,7.5 "X39","D-Xylitol",0.429398587022026,0.637172096871394,0.296217775741559,1.3 "X40","D-Arabitol",0.157959973710896,0.619253978472208,0.28788774182491,1.8 "X41","Ribitol",0.604761035100845,0.793505619284379,0.368896363693543,1.2 "X42","L-Fucose",0.670189357605648,0.825695968188865,0.383861478455554,1.4 "X43","cis-Aconitic acid",0.110228853838941,0.619253978472208,0.28788774182491,1.4 "X44","D-Ribose",0.941147698007397,0.96206209129645,0.447257333157555,0.9 "X45","Homovanillic acid",0.0210598598317868,0.619253978472208,0.28788774182491,3.8 "X46","Glycerol 3-phosphate",0.862835879992945,0.912424148958056,0.42418092892695,0.9 "X47","L-Glutamine",0.327213751112235,0.619253978472208,0.28788774182491,3 "X48","",0.907768597469626,0.949030806445518,0.441199161067862,0.9 "X49","Hypoxanthine",0.151292577387652,0.619253978472208,0.28788774182491,2.7 "X50","Hippuric acid",0.334740288033347,0.619253978472208,0.28788774182491,1.5 "X51","Ornithine",0.329256577171709,0.619253978472208,0.28788774182491,0 "X52","Citric acid",0.386365212751301,0.619253978472208,0.28788774182491,1.2 "X53","3-(3-Hydroxyphenyl)-3-hydroxypropanoic acid",0.168655111865478,0.619253978472208,0.28788774182491,2.2 "X54","",0.293404566764687,0.619253978472208,0.28788774182491,0.7 "X55","Quinic acid",0.37146956942578,0.619253978472208,0.28788774182491,0.1 "X56","L-Tyrosine",0.130286001429568,0.619253978472208,0.28788774182491,1.6 "X57","L-Sorbose",0.728005123544874,0.84824624795081,0.394344978511868,1.2 "X58","D-Fructose",0.40687608994573,0.619253978472208,0.28788774182491,0.3 "X59","D-Galactose",0.23954691170069,0.619253978472208,0.28788774182491,10.8 "X60","D-Glucose",0.527272928664007,0.724016558762518,0.33659128466195,1.4 "X61","L-Histidine",0.231586291516813,0.619253978472208,0.28788774182491,2.6 "X62","Mannitol",0.410592311813094,0.619253978472208,0.28788774182491,1.6 "X63","",0.132368316110587,0.619253978472208,0.28788774182491,2.2 "X64","Sorbitol",0.704670638158063,0.84824624795081,0.394344978511868,0.6 "X65","",0.826738024894663,0.884417421980337,0.411160756807241,1.4 "X66","Galactitol",0.362332614944998,0.619253978472208,0.28788774182491,0.5 "X67","",0.403383482031547,0.619253978472208,0.28788774182491,0.5 "X68","Gulonic acid",0.283817109684353,0.619253978472208,0.28788774182491,3.1 "X69","Gluconic acid",0.756449180307697,0.849001881328481,0.394696268280376,1.2 "X70","",0.47195325607601,0.667995377830661,0.310547348194029,1.2 "X71","Kynurenic acid",0.974993982358594,0.985708201944952,0.458250278918411,1 "X72","Beta-N-Acetylglucosamine",0.625895495671477,0.793505619284379,0.368896363693543,0.6 "X73","",0.775175630778178,0.849001881328481,0.394696268280376,0.9 "X74","Uric acid",0.723484270841933,0.84824624795081,0.394344978511868,0.9 "X75","",0.364447095492839,0.619253978472208,0.28788774182491,1.6 "X76","",0.617439420550046,0.793505619284379,0.368896363693543,0.8 "X77","L-Tryptophan",0.0463971338065384,0.619253978472208,0.28788774182491,3.5 "X78","5-Hydroxyindoleacetic acid",0.168074250813843,0.619253978472208,0.28788774182491,19 "X79","",0.373720389297531,0.619253978472208,0.28788774182491,0.6 "X80","4-Hydroxyhippuric acid",0.356128215300218,0.619253978472208,0.28788774182491,1.4 "X81","",0.329256577171709,0.619253978472208,0.28788774182491,0 "X82","L-Cystine",0.325981442466079,0.619253978472208,0.28788774182491,0 "X83","",0.391209618293358,0.619253978472208,0.28788774182491,0.6 "X84","Pseudouridine",0.673121713197445,0.825695968188865,0.383861478455554,1.1 "X85","Arabinofuranose",0.0979348834661696,0.619253978472208,0.28788774182491,1.4 "X86","",0.307117130523681,0.619253978472208,0.28788774182491,1.4 "X87","Acetaminophen glucuronide",0.629629458779996,0.793505619284379,0.368896363693543,0.8 "X88","Hydroxyproline dipeptide 4-TMS",0.269089443348433,0.619253978472208,0.28788774182491,4.1 "X89","Xylobiose",0.192736164387755,0.619253978472208,0.28788774182491,1.8 "X90","Sucrose",0.393758561766684,0.619253978472208,0.28788774182491,1.8 "X91","",0.181333607978092,0.619253978472208,0.28788774182491,50 "X92","Alpha-Lactose",0.511447992874468,0.712927505218956,0.331436045140505,1.5 ================================================ FILE: Demos/Mapped Network From Data (Biochemical and Structural)/variable info.csv ================================================ ID,name,Name2,HMDB,PubChem,KEGG 1,1-2-Propandiol,Propylene glycol,HMDB01881,1030,C00583 2,Glycolic acid,Glycolic acid,HMDB00115,757,C00160 3,Alanine 2-TMS,L-Alanine,HMDB00161,5950,C00041 4,S,,,, 5,Valine 2-TMS,L-Valine,HMDB00883,6287,C00183 6,Urea,Urea,HMDB00294,1176,C00086 7,Benzoic acid,Benzoic acid,HMDB01870,243,C00180 8,Serine,L-Serine,HMDB00187,5951,C00065 9,Ethanolamine,Ethanolamine,HMDB00149,700,C00189 10,Phosphate,Phosphoric acid,HMDB02142,1004,C00009 11,Threonine,L-Threonine,HMDB00167,6288,C00188 12,Glycine,Glycine,HMDB00123,750,C00037 13,Succinic acid,Succinic acid,HMDB00254,1110,C00042 14,2-3-Dihydroxybutyric acid,Glyceric acid,HMDB00139,439194,C00258 15,2-4-Dihydroxybutyric acid,"2,4-Dihydroxybutanoic acid",HMDB00360,192742, 16,3-4-Dihydroxybutyric acid,"(S)-3,4-Dihydroxybutyric acid",HMDB00337,150929, 17,3-Aminoisobutyric acid 3-TMS,3-Aminoisobutanoic acid,HMDB03911,64956,C05145 18,Aminomalonic acid,Aminomalonic acid,HMDB01147,100714,C00872 19,1-2-3-Butanetriol,"(2R*,3R*)-1,2,3-Butanetriol",HMDB34778,20497, 20,Threitol,D-Threitol,HMDB04136,169019,C16884 21,Erythritol,Erythritol,HMDB02994,222285,C00503 22,Pyroglutamic acid,Pyroglutamic acid,HMDB00267,7405,C01879 23,1-Deoxypentitol,,,, 24,Erythropentose-2-deoxy A- 2-Deoxy-ribose,,,, 25,Erythronic or threonic acid,Erythronic acid,HMDB00613,2781043, 26,Cysteine 3-TMS,L-Cysteine,HMDB00574,5862,C00097 27,Creatinine 3-TMS,Creatinine,HMDB00562,588,C00791 28,2-3-4-Trihydroxybutyric acid (Threonic or erythronic),Threonic acid,HMDB00943,151152,C01620 29,2-Ketoglutaric acid,Oxoglutaric acid,HMDB00208,51,C00026 30,Phenylalanine 2-TMS,L-Phenylalanine,HMDB00159,6140,C00079 31,3-Deoxypentitol,,,, 32,4-Hydroxyphenylacetic acid,p-Hydroxyphenylacetic acid,HMDB00020,127,C00642 33,3-4-5-Trihydroxypentanoic acid,,,, 34,Taurine 3-TMS,Taurine,HMDB00251,1123,C00245 35,Lyxose A,Pectin,HMDB03402,441476,C08348 36,Arabinose,Arabinose,HMDB29942,66308, 37,Ribose A,D-Ribose,HMDB00283,5779,C00121 38,Lysine,L-Lysine,HMDB00182,5962,C00047 39,Xylitol,D-Xylitol,HMDB02917,6912,C00379 40,Arabitol,D-Arabitol,HMDB00568,827,C01904 41,Ribitol (Adonitol),Ribitol,HMDB00508, ,C00474 42,Fucose (6-Deoxygalactose),L-Fucose,HMDB00174,17106,C01019 43,cis-Aconitic acid,cis-Aconitic acid,HMDB00072,643757,C00417 44,Ribose 4-TMS,D-Ribose,HMDB00283,5779,C00121 45,Homovanillic acid,Homovanillic acid,HMDB00118,1738,C05582 46,Glycerol-3-phosphate,Glycerol 3-phosphate,HMDB00126,439162,C00093 47,Glutamine 3-TMS,L-Glutamine,HMDB00641,5961,C00064 48,2-2-Dimethyl-3-hydroxybutyric acid,,,, 49,Hypoxanthine,Hypoxanthine,HMDB00157,790,C00262 50,Hippuric acid,Hippuric acid,HMDB00714,464,C01586 51,Ornithine 3-TMS,Ornithine,HMDB00214,6262,C00077 52,Citric acid,Citric acid,HMDB00094,311,C00158 53,3-(3-Hydroxyphenyl)-3-hydroxypropionic acid 3-TMS,3-(3-Hydroxyphenyl)-3-hydroxypropanoic acid,HMDB02643,102959, 54,Glucoheptonic acid 1-4-lactone,,,, 55,Quinic acid,Quinic acid,HMDB03072, ,C06746 56,Tyrosine,L-Tyrosine,HMDB00158,6057,C00082 57,Sorbose,L-Sorbose,HMDB01266,441484,C08356 58,Fructose,D-Fructose,HMDB00660,439709,C02336 59,Galactose,D-Galactose,HMDB00143,439357,C00984 60,Glucose,D-Glucose,HMDB00122,5793,C00031 61,Histidine 3-TMS,L-Histidine,HMDB00177,6274,C00135 62,Mannitol,Mannitol,HMDB00765,6251,C00392 63,2-3-4-5-Tetrahydroxypentanoic acid 1-4-lactone,,,, 64,Sorbitol,Sorbitol,HMDB00247,5780,C00794 65,Isoascorbic acid,,,, 66,Galactitol,Galactitol,HMDB00107,11850,C01697 67,2-O-Glycerol-galactopyranoside 6-TMS,,,, 68,Gulonic acid,Gulonic acid,HMDB03290,152304,C00800 69,Gluconic acid,Gluconic acid,HMDB00625,10690,C00257 70,Inositol-like,,,, 71,Kynurenic acid,Kynurenic acid,HMDB00715,3845,C01717 72,N-Acetylglucosamine,Beta-N-Acetylglucosamine,HMDB00803,24139,C03878 73,Inositol,,,, 74,Uric acid,Uric acid,HMDB00289,1175,C00366 75,Glycine- N-(4-Hydroxybenzoyl) 2-TMS deriv.,,,, 76,E,,,, 77,Tryptophan 2-TMS,L-Tryptophan,HMDB00929,6305,C00078 78,5-Hydroxyindole-3-acetic acid (5-HIAA),5-Hydroxyindoleacetic acid,HMDB00763,1826,C05635 79,C,,,, 80,Glycine- N-(4-Hydroxybenzoyl) 2-TMS,4-Hydroxyhippuric acid,HMDB13678,151012, 81,F,,,, 82,Cystine 4-TMS,L-Cystine,HMDB00192,67678,C00491 83,2-O-Glycerol-galactopyranoside 6-TMS,,,, 84,Pseudouridine,Pseudouridine,HMDB00767,15047,C02067 85,Arabinofuranose- 1-2-3-4-OTMS,Arabinofuranose,HMDB12325,440921,C06115 86,A,,,, 87,Paracetamol glucuronide,Acetaminophen glucuronide,HMDB10316,83944, 88,Hydroxyproline dipeptide 4-TMS,Hydroxyproline dipeptide 4-TMS, , , 89,Xylobiose-6-TMS,Xylobiose,HMDB29894, ,C01630 90,Sucrose,Sucrose,HMDB00258,5988,C00089 91,B,,,, 92,Lactose,Alpha-Lactose,HMDB00186,84571,C00243 ================================================ FILE: Demos/Pathway Analysis/.xml ================================================ ================================================ FILE: Demos/Pathway Analysis/KEGG Pathway Enrichment.Rmd ================================================ Visualization of KEGG Pathway Enrichment ======================================================== ![](http://pathview.r-forge.r-project.org/images/hsa01100.sim.cpd.png) The [KEGG](http://www.genome.jp/kegg/) database is a useful repository of biochemical domain knowledge. The following is an example of how to map changes in genes, proteins and metabolites on an organism specific basis to KEGG defined [biochemical pathways](http://www.genome.jp/kegg/pathway.html). For this example we will use the R packages [pathview](http://bioconductor.org/packages/release/bioc/html/pathview.html), [KEGGREST](http://bioconductor.org/packages/2.12/bioc/html/KEGGREST.html) and [KEGGgraph](http://www.bioconductor.org/packages/release/bioc/html/KEGGgraph.html). To generate a pathway enrichment visualization we minimally need a list of fold changes in biochemical components of interest (e.g. genes, proteins, metabolites). We need to supply gene/protein and metabolite/compound fold changes separately as data.frames with rownames specifying database identifiers (see [pathview](http://pathview.r-forge.r-project.org/) for possible options). This can be generated based on the output from statistical analysis used to identify any significantly differentially expressed species. Lets create gene and metabolite data for a demonstration. ```{r,message=FALSE,warning=FALSE} library(pathview) metabolite.data<- data.frame(FC=sim.mol.data(mol.type="cpd", nmol=3000)) ``` Lets take a look at the format of the metabolite inputs. ```{r} head(metabolite.data) ``` Notice we have a single column data frame containing log fold changes with KEGG identifiers for the rownames. Next lets load the example gene data. For genes we can specify the organism of interest using the ```r 'species'``` argument. We can check for available organisms using the commands below. ```{r} data(korg) head(korg) ``` We can use the columns ```r 'scientific.name'``` or ```r 'common.name'``` to search for the ```r "kegg.code"``` for our organism of interest. Here is an example of how we can do this for *arabidopsis thaliana*. ```{r} organism<-"arabidopsis thaliana" matches<-unlist(sapply(1:ncol(korg),function(i) {agrep(organism,korg[,i])})) (kegg.code<-korg[matches,1,drop=F]) ``` ```{r} #load gene data gene.data<- data.frame(FC=sim.mol.data(mol.type="gene", nmol=3000,species=kegg.code)) ``` Lets take a look at the gene data. ```{r} head(gene.data) ``` Looking at the rownames we can see that the identifiers are specific for arabidopsis thaliana. Using real data it is possible that we will need to translate identifiers to match the type used in pathview. Here are the different identifiers which can be supplied for metabolites. ```{r} data(cpd.simtypes) cpd.simtypes ``` Here are some possible options for gene identifiers. ```{r} data(gene.idtype.list) gene.idtype.list ``` We can optionally simulate data with with other identifiers by selecting the appropriate identifier type for the argument `id.type` in the function ```r "sim.mol.data"``` . In the example data above we have the the logarithm of fold changes for `r nrow(metabolite.data)` metabolites and `r nrow(gene.data)` genes. Now that we have the data we can select some pathway of interest to map the fold changes to. We could get this information from a pathway enrichment analysis. We would do this by testing if the significantly differential expressed metabolites or genes are enriched for some specific KEGG pathways of interest. Here are some tools to conduct enrichment analysis for genes and metabolites. * [MBrole](http://csbg.cnb.csic.es/mbrole/) (metabolites) * [MetaboAnalyst](http://www.metaboanalyst.ca/MetaboAnalyst/faces/Home.jsp) (metabolites) * [David](http://david.abcc.ncifcrf.gov/) (genes) * [IMPaLA](http://impala.molgen.mpg.de/) (genes and metabolites) For the example below we will randomly select some KEGG pathways. We can use the R package [KEGGREST](http://bioconductor.org/packages/2.12/bioc/html/KEGGREST.html) to get all KEGG pathway identifiers for ```r organism ```. ```{r,message=FALSE,warning=FALSE} #get names of pathways to visualize library(KEGGREST) pathways<-keggList("pathway" ,kegg.code) head(pathways) ``` Lets visualize changes in our genes and metabolites for ```r pathways[2] ``` which has the KEGG id ```r names(pathways)[2] ``` . ```{r,message=FALSE,warning=FALSE} library(pathview) map<-gsub("path:","",names(pathways)[2]) # remove 'path:' pv.out <- pathview(gene.data = gene.data, cpd.data = metabolite.data, gene.idtype = "KEGG", pathway.id = map, species = kegg.code, out.suffix = map, keys.align = "y", kegg.native = T, match.data=T, key.pos = "topright") plot.name<-paste(map,map,"png",sep=".") ``` If everything went as planned this generated a file named ```r plot.name``` mapped KEGG pathway. ![ath00020.ath00020.png](ath00020.ath00020.png) We can take a look at the mappings made to this pathway. ```{r} head(pv.out) ``` We can also display changes in genes/proteins and metabolites for multiple comparisons. To do this lets create some more artificial data and this time lets get protein information. ```{r} metabolite.data2<- sim.mol.data(mol.type="cpd", nmol=3000,nexp=2) head(metabolite.data2) gene.data2<- sim.mol.data(mol.type="gene", nmol=3000,nexp=2,id.type="UNIPROT") head(gene.data2) ``` Notice UNIPROT IDs do not map to ```r organism ```. We can check the available mappings by looking at the database of identifiers for ```r organism ``` org.At.tair.db. Lets instead map changes to a human pathway or KEGG code "hsa". To do this we need to make sure we supply the correct ```r 'species' ``` and ```r 'gene.idtype'``` arguments. Next lets map changes in proteins and metabolites for the two comparisons to the pathway . ```{r, message=FALSE,warning=FALSE} map<-gsub("path:ath","",names(pathways)[1]) # remove 'path:ath' pv.out <- pathview(gene.data = gene.data2, cpd.data = metabolite.data2, gene.idtype = "UNIPROT", pathway.id = map, species = "hsa", out.suffix = map, keys.align = "y", kegg.native = T, match.data=T, key.pos = "topright") plot.name<-paste(map,map,"multi","png",sep=".") ``` This should have generated an image file named ```r plot.name```. ![hsa00010.00010.multi.png](hsa00010.00010.multi.png) We could also map metabolites on a large scale using `pathway.id= '01100'`. ![metabolic pathways](http://pathview.r-forge.r-project.org/images/hsa01100.sim.cpd.png) ```{r} sessionInfo() ``` © Dmitry Grapov (2014) Creative Commons License ================================================ FILE: Demos/Pathway Analysis/KEGG Pathway Enrichment.html ================================================ Visualization of KEGG Pathway Enrichment

Visualization of KEGG Pathway Enrichment

The KEGG database is a useful repository of biochemical domain knowledge. The following is an example of how to map changes in genes, proteins and metabolites on an organism specific basis to KEGG defined biochemical pathways. For this example we will use the R packages pathview, KEGGREST and KEGGgraph.

To generate a pathway enrichment visualization we minimally need a list of fold changes in biochemical components of interest (e.g. genes, proteins, metabolites). We need to supply gene/protein and metabolite/compound fold changes separately as data.frames with rownames specifying database identifiers (see pathview for possible options). This can be generated based on the output from statistical analysis used to identify any significantly differentially expressed species.

Lets create gene and metabolite data for a demonstration.

library(pathview)
metabolite.data <- data.frame(FC = sim.mol.data(mol.type = "cpd", nmol = 3000))

Lets take a look at the format of the metabolite inputs. 

head(metabolite.data)

##              FC
## C02787 -1.15260
## C08521  0.46416
## C01043  0.72893
## C11496  0.41062
## C07111 -1.46115
## C00031 -0.01891

Notice we have a single column data frame containing log fold changes with KEGG identifiers for the rownames. Next lets load the example gene data. For genes we can specify the organism of interest using the species argument. We can check for available organisms using the commands below.

data(korg)
head(korg)

##      kegg.code scientific.name           common.name              
## [1,] "hsa"     "Homo sapiens"            "human"                  
## [2,] "ptr"     "Pan troglodytes"         "chimpanzee"             
## [3,] "pps"     "Pan paniscus"            "bonobo"                 
## [4,] "ggo"     "Gorilla gorilla gorilla" "western lowland gorilla"
## [5,] "pon"     "Pongo abelii"            "Sumatran orangutan"     
## [6,] "mcc"     "Macaca mulatta"          "rhesus monkey"          
##      entrez.gnodes kegg.geneid ncbi.geneid
## [1,] "1"           "100"       "100"      
## [2,] "1"           "100533953" "100533953"
## [3,] "1"           "100967419" "100967419"
## [4,] "1"           "101123859" "101123859"
## [5,] "1"           "100169736" "100169736"
## [6,] "1"           "100301991" "100301991"

We can use the columns scientific.name or common.name to search for the kegg.code for our organism of interest. Here is an example of how we can do this for arabidopsis thaliana.

organism <- "arabidopsis thaliana"
matches <- unlist(sapply(1:ncol(korg), function(i) {
    agrep(organism, korg[, i])
}))
(kegg.code <- korg[matches, 1, drop = F])

##      kegg.code
## [1,] "ath"

# load gene data
gene.data <- data.frame(FC = sim.mol.data(mol.type = "gene", nmol = 3000, species = kegg.code))

Lets take a look at the gene data.

head(gene.data)

##                 FC
## AT2G21190 -1.15260
## AT1G80990  0.46416
## AT3G48550  0.72893
## AT1G15410  0.41062
## AT3G15220 -1.46115
## AT3G18980 -0.01891

Looking at the rownames we can see that the identifiers are specific for arabidopsis thaliana. Using real data it is possible that we will need to translate identifiers to match the type used in pathview. Here are the different identifiers which can be supplied for metabolites.

data(cpd.simtypes)
cpd.simtypes

## [1] "Beilstein Registry Number" "CAS Registry Number"      
## [3] "ChEMBL COMPOUND"           "KEGG COMPOUND accession"  
## [5] "KEGG DRUG accession"       "Patent accession"         
## [7] "PubMed citation"

Here are some possible options for gene identifiers.

data(gene.idtype.list)
gene.idtype.list

##  [1] "SYMBOL"       "GENENAME"     "ENSEMBL"      "ENSEMBLPROT" 
##  [5] "PROSITE"      "UNIGENE"      "UNIPROT"      "ACCNUM"      
##  [9] "ENSEMBLTRANS" "REFSEQ"

We can optionally simulate data with with other identifiers by selecting the appropriate identifier type for the argument id.type in the function sim.mol.data .

In the example data above we have the the logarithm of fold changes for 3000 metabolites and 3000 genes. Now that we have the data we can select some pathway of interest to map the fold changes to. We could get this information from a pathway enrichment analysis. We would do this by testing if the significantly differential expressed metabolites or genes are enriched for some specific KEGG pathways of interest. Here are some tools to conduct enrichment analysis for genes and metabolites.

For the example below we will randomly select some KEGG pathways. We can use the R package KEGGREST to get all KEGG pathway identifiers for arabidopsis thaliana.

# get names of pathways to visualize
library(KEGGREST)
pathways <- keggList("pathway", kegg.code)
head(pathways)

##                                                                   path:ath00010 
##             "Glycolysis / Gluconeogenesis - Arabidopsis thaliana (thale cress)" 
##                                                                   path:ath00020 
##                "Citrate cycle (TCA cycle) - Arabidopsis thaliana (thale cress)" 
##                                                                   path:ath00030 
##                "Pentose phosphate pathway - Arabidopsis thaliana (thale cress)" 
##                                                                   path:ath00040 
## "Pentose and glucuronate interconversions - Arabidopsis thaliana (thale cress)" 
##                                                                   path:ath00051 
##          "Fructose and mannose metabolism - Arabidopsis thaliana (thale cress)" 
##                                                                   path:ath00052 
##                     "Galactose metabolism - Arabidopsis thaliana (thale cress)"

Lets visualize changes in our genes and metabolites for Citrate cycle (TCA cycle) - Arabidopsis thaliana (thale cress) which has the KEGG id path:ath00020 .

library(pathview)
map <- gsub("path:", "", names(pathways)[2])  # remove 'path:'
pv.out <- pathview(gene.data = gene.data, cpd.data = metabolite.data, gene.idtype = "KEGG", 
    pathway.id = map, species = kegg.code, out.suffix = map, keys.align = "y", 
    kegg.native = T, match.data = T, key.pos = "topright")
plot.name <- paste(map, map, "png", sep = ".")

If everything went as planned this generated a file named ath00020.ath00020.png mapped KEGG pathway. ath00020.ath00020.png We can take a look at the mappings made to this pathway.

head(pv.out)

## $plot.data.gene
##    kegg.names       labels type   x   y width height      FC mol.col
## 29  AT1G48030       mtLPD1 gene 467 618    46     17      NA #FFFFFF
## 30  AT3G55410    AT3G55410 gene 661 574    46     17      NA #FFFFFF
## 31  AT3G55410    AT3G55410 gene 530 575    46     17      NA #FFFFFF
## 32  AT4G26910    AT4G26910 gene 403 574    46     17      NA #FFFFFF
## 33  AT2G20420    AT2G20420 gene 260 574    46     17      NA #FFFFFF
## 34  AT2G20420    AT2G20420 gene 260 555    46     17      NA #FFFFFF
## 35  AT1G54340         ICDH gene 718 505    46     17      NA #FFFFFF
## 36  AT5G03290       IDH-IV gene 766 458    46     17 -0.4385 #5FDF5F
## 37  AT1G54340         ICDH gene 718 400    46     17      NA #FFFFFF
## 38  AT1G10670       ACLA-3 gene 434 355    46     17 -1.1941 #00FF00
## 39  AT2G47510         FUM1 gene 191 436    46     17      NA #FFFFFF
## 40  AT2G05710         ACO3 gene 670 345    46     17      NA #FFFFFF
## 41  AT2G05710         ACO3 gene 571 345    46     17      NA #FFFFFF
## 42  AT2G42790         CSY3 gene 434 334    46     17      NA #FFFFFF
## 43  AT1G04410    AT1G04410 gene 253 344    46     17      NA #FFFFFF
## 77  AT1G30120 PDH-E1_ALPHA gene 686 241    46     17  0.6117 #EF3030
## 79  AT1G30120 PDH-E1_ALPHA gene 582 241    46     17  0.6117 #EF3030
## 82  AT1G48030       mtLPD1 gene 529 285    46     17      NA #FFFFFF
## 84  AT1G34430      EMB3003 gene 464 242    46     17      NA #FFFFFF
## 91  AT4G37870         PCK1 gene 371 138    46     17      NA #FFFFFF
## 97  AT5G66760       SDH1-2 gene 191 530    46     17  1.1329 #FF0000
## 
## $plot.data.cpd
##    kegg.names labels     type   x   y width height       FC mol.col
## 58     C00022 C00022 compound 737 241     8      8  0.21082 #CECE8F
## 59     C00122 C00122 compound 190 481     8      8       NA #FFFFFF
## 60     C00036 C00036 compound 320 344     8      8  1.42760 #FFFF00
## 61     C05379 C05379 compound 718 453     8      8  0.76991 #EFEF30
## 62     C00024 C00024 compound 400 241     8      8       NA #FFFFFF
## 63     C00149 C00149 compound 190 393     8      8  1.93744 #FFFF00
## 64     C00311 C00311 compound 718 344     8      8 -1.33400 #0000FF
## 65     C00417 C00417 compound 621 344     8      8 -0.82811 #0000FF
## 66     C00042 C00042 compound 190 574     8      8  1.28575 #FFFF00
## 67     C00158 C00158 compound 522 344     8      8       NA #FFFFFF
## 68     C15972 C15972 compound 522 617     8      8       NA #FFFFFF
## 69     C00068 C00068 compound 589 530     8      8  0.01167 #BEBEBE
## 70     C16254 C16254 compound 462 573     8      8       NA #FFFFFF
## 71     C15973 C15973 compound 408 617     8      8       NA #FFFFFF
## 72     C00091 C00091 compound 334 574     8      8 -1.97370 #0000FF
## 73     C00026 C00026 compound 718 573     8      8  1.10432 #FFFF00
## 74     C05381 C05381 compound 593 573     8      8       NA #FFFFFF
## 76     C05125 C05125 compound 631 241     8      8       NA #FFFFFF
## 78     C00068 C00068 compound 631 198     8      8  0.01167 #BEBEBE
## 80     C15972 C15972 compound 580 284     8      8       NA #FFFFFF
## 83     C16255 C16255 compound 517 241     8      8       NA #FFFFFF
## 93     C15973 C15973 compound 478 284     8      8       NA #FFFFFF
## 94     C00074 C00074 compound 471 127     8      8       NA #FFFFFF

We can also display changes in genes/proteins and metabolites for multiple comparisons. To do this lets create some more artificial data and this time lets get protein information. 

metabolite.data2 <- sim.mol.data(mol.type = "cpd", nmol = 3000, nexp = 2)
head(metabolite.data2)

##            exp1    exp2
## C02787 -1.15260  1.2914
## C08521  0.46416  0.6706
## C01043  0.72893 -0.3065
## C11496  0.41062 -2.1996
## C07111 -1.46115 -0.3476
## C00031 -0.01891  0.7219

gene.data2 <- sim.mol.data(mol.type = "gene", nmol = 3000, nexp = 2, id.type = "UNIPROT")
head(gene.data2)

##            exp1    exp2
## E9PDR7 -1.15260  1.2914
## Q9UBV2  0.46416  0.6706
## Q9BQ95  0.72893 -0.3065
## P24310  0.41062 -2.1996
## Q8N5I0 -1.46115 -0.3476
## Q9Y6J0 -0.01891  0.7219

Notice UNIPROT IDs do not map to arabidopsis thaliana. We can check the available mappings by looking at the database of identifiers for arabidopsis thaliana org.At.tair.db. Lets instead map changes to a human pathway or KEGG code “hsa”. To do this we need to make sure we supply the correct species and gene.idtype arguments. Next lets map changes in proteins and metabolites for the two comparisons to the pathway .

map <- gsub("path:ath", "", names(pathways)[1])  # remove 'path:ath'
pv.out <- pathview(gene.data = gene.data2, cpd.data = metabolite.data2, gene.idtype = "UNIPROT", 
    pathway.id = map, species = "hsa", out.suffix = map, keys.align = "y", kegg.native = T, 
    match.data = T, key.pos = "topright")
plot.name <- paste(map, map, "multi", "png", sep = ".")

This should have generated an image file named 00010.00010.multi.png. hsa00010.00010.multi.png

We could also map metabolites on a large scale using pathway.id= '01100'. metabolic pathways

sessionInfo()

## R version 3.0.1 (2013-05-16)
## Platform: i386-w64-mingw32/i386 (32-bit)
## 
## locale:
## [1] LC_COLLATE=English_United States.1252 
## [2] LC_CTYPE=English_United States.1252   
## [3] LC_MONETARY=English_United States.1252
## [4] LC_NUMERIC=C                          
## [5] LC_TIME=English_United States.1252    
## 
## attached base packages:
## [1] parallel  stats     graphics  grDevices utils     datasets  methods  
## [8] base     
## 
## other attached packages:
##  [1] KEGGREST_1.2.2       pathview_1.2.4       org.Hs.eg.db_2.9.0  
##  [4] RSQLite_0.11.3       DBI_0.2-7            AnnotationDbi_1.24.0
##  [7] Biobase_2.22.0       BiocGenerics_0.8.0   KEGGgraph_1.16.0    
## [10] graph_1.40.1         XML_3.96-1.1         knitr_1.5           
## [13] rcom_2.2-5           rscproxy_2.0-5      
## 
## loaded via a namespace (and not attached):
##  [1] Biostrings_2.30.1 evaluate_0.5.3    formatR_0.10     
##  [4] grid_3.0.1        httr_0.3          IRanges_1.20.7   
##  [7] png_0.1-7         RCurl_1.95-4.1    Rgraphviz_2.6.0  
## [10] stats4_3.0.1      stringr_0.6.2     tools_3.0.1      
## [13] XVector_0.2.0

© Dmitry Grapov (2014) Creative Commons License

================================================ FILE: Demos/Pathway Analysis/KEGG Pathway Enrichment.md ================================================ Visualization of KEGG Pathway Enrichment ======================================================== ![](http://pathview.r-forge.r-project.org/images/hsa01100.sim.cpd.png) The [KEGG](http://www.genome.jp/kegg/) database is a useful repository of biochemical domain knowledge. The following is an example of how to map changes in genes, proteins and metabolites on an organism specific basis to KEGG defined [biochemical pathways](http://www.genome.jp/kegg/pathway.html). For this example we will use the R packages [pathview](http://bioconductor.org/packages/release/bioc/html/pathview.html), [KEGGREST](http://bioconductor.org/packages/2.12/bioc/html/KEGGREST.html) and [KEGGgraph](http://www.bioconductor.org/packages/release/bioc/html/KEGGgraph.html). To generate a pathway enrichment visualization we minimally need a list of fold changes in biochemical components of interest (e.g. genes, proteins, metabolites). We need to supply gene/protein and metabolite/compound fold changes separately as data.frames with rownames specifying database identifiers (see [pathview](http://pathview.r-forge.r-project.org/) for possible options). This can be generated based on the output from statistical analysis used to identify any significantly differentially expressed species. Lets create gene and metabolite data for a demonstration. ```r library(pathview) metabolite.data <- data.frame(FC = sim.mol.data(mol.type = "cpd", nmol = 3000)) ``` Lets take a look at the format of the metabolite inputs. ```r head(metabolite.data) ``` ``` ## FC ## C02787 -1.15260 ## C08521 0.46416 ## C01043 0.72893 ## C11496 0.41062 ## C07111 -1.46115 ## C00031 -0.01891 ``` Notice we have a single column data frame containing log fold changes with KEGG identifiers for the rownames. Next lets load the example gene data. For genes we can specify the organism of interest using the ``species`` argument. We can check for available organisms using the commands below. ```r data(korg) head(korg) ``` ``` ## kegg.code scientific.name common.name ## [1,] "hsa" "Homo sapiens" "human" ## [2,] "ptr" "Pan troglodytes" "chimpanzee" ## [3,] "pps" "Pan paniscus" "bonobo" ## [4,] "ggo" "Gorilla gorilla gorilla" "western lowland gorilla" ## [5,] "pon" "Pongo abelii" "Sumatran orangutan" ## [6,] "mcc" "Macaca mulatta" "rhesus monkey" ## entrez.gnodes kegg.geneid ncbi.geneid ## [1,] "1" "100" "100" ## [2,] "1" "100533953" "100533953" ## [3,] "1" "100967419" "100967419" ## [4,] "1" "101123859" "101123859" ## [5,] "1" "100169736" "100169736" ## [6,] "1" "100301991" "100301991" ``` We can use the columns ``scientific.name`` or ``common.name`` to search for the ``kegg.code`` for our organism of interest. Here is an example of how we can do this for *arabidopsis thaliana*. ```r organism <- "arabidopsis thaliana" matches <- unlist(sapply(1:ncol(korg), function(i) { agrep(organism, korg[, i]) })) (kegg.code <- korg[matches, 1, drop = F]) ``` ``` ## kegg.code ## [1,] "ath" ``` ```r # load gene data gene.data <- data.frame(FC = sim.mol.data(mol.type = "gene", nmol = 3000, species = kegg.code)) ``` Lets take a look at the gene data. ```r head(gene.data) ``` ``` ## FC ## AT2G21190 -1.15260 ## AT1G80990 0.46416 ## AT3G48550 0.72893 ## AT1G15410 0.41062 ## AT3G15220 -1.46115 ## AT3G18980 -0.01891 ``` Looking at the rownames we can see that the identifiers are specific for arabidopsis thaliana. Using real data it is possible that we will need to translate identifiers to match the type used in pathview. Here are the different identifiers which can be supplied for metabolites. ```r data(cpd.simtypes) cpd.simtypes ``` ``` ## [1] "Beilstein Registry Number" "CAS Registry Number" ## [3] "ChEMBL COMPOUND" "KEGG COMPOUND accession" ## [5] "KEGG DRUG accession" "Patent accession" ## [7] "PubMed citation" ``` Here are some possible options for gene identifiers. ```r data(gene.idtype.list) gene.idtype.list ``` ``` ## [1] "SYMBOL" "GENENAME" "ENSEMBL" "ENSEMBLPROT" ## [5] "PROSITE" "UNIGENE" "UNIPROT" "ACCNUM" ## [9] "ENSEMBLTRANS" "REFSEQ" ``` We can optionally simulate data with with other identifiers by selecting the appropriate identifier type for the argument `id.type` in the function ``sim.mol.data`` . In the example data above we have the the logarithm of fold changes for 3000 metabolites and 3000 genes. Now that we have the data we can select some pathway of interest to map the fold changes to. We could get this information from a pathway enrichment analysis. We would do this by testing if the significantly differential expressed metabolites or genes are enriched for some specific KEGG pathways of interest. Here are some tools to conduct enrichment analysis for genes and metabolites. * [MBrole](http://csbg.cnb.csic.es/mbrole/) (metabolites) * [MetaboAnalyst](http://www.metaboanalyst.ca/MetaboAnalyst/faces/Home.jsp) (metabolites) * [David](http://david.abcc.ncifcrf.gov/) (genes) * [IMPaLA](http://impala.molgen.mpg.de/) (genes and metabolites) For the example below we will randomly select some KEGG pathways. We can use the R package [KEGGREST](http://bioconductor.org/packages/2.12/bioc/html/KEGGREST.html) to get all KEGG pathway identifiers for ``arabidopsis thaliana``. ```r # get names of pathways to visualize library(KEGGREST) pathways <- keggList("pathway", kegg.code) head(pathways) ``` ``` ## path:ath00010 ## "Glycolysis / Gluconeogenesis - Arabidopsis thaliana (thale cress)" ## path:ath00020 ## "Citrate cycle (TCA cycle) - Arabidopsis thaliana (thale cress)" ## path:ath00030 ## "Pentose phosphate pathway - Arabidopsis thaliana (thale cress)" ## path:ath00040 ## "Pentose and glucuronate interconversions - Arabidopsis thaliana (thale cress)" ## path:ath00051 ## "Fructose and mannose metabolism - Arabidopsis thaliana (thale cress)" ## path:ath00052 ## "Galactose metabolism - Arabidopsis thaliana (thale cress)" ``` Lets visualize changes in our genes and metabolites for ``Citrate cycle (TCA cycle) - Arabidopsis thaliana (thale cress)`` which has the KEGG id ``path:ath00020`` . ```r library(pathview) map <- gsub("path:", "", names(pathways)[2]) # remove 'path:' pv.out <- pathview(gene.data = gene.data, cpd.data = metabolite.data, gene.idtype = "KEGG", pathway.id = map, species = kegg.code, out.suffix = map, keys.align = "y", kegg.native = T, match.data = T, key.pos = "topright") plot.name <- paste(map, map, "png", sep = ".") ``` If everything went as planned this generated a file named ``ath00020.ath00020.png`` mapped KEGG pathway. ![ath00020.ath00020.png](ath00020.ath00020.png) We can take a look at the mappings made to this pathway. ```r head(pv.out) ``` ``` ## $plot.data.gene ## kegg.names labels type x y width height FC mol.col ## 29 AT1G48030 mtLPD1 gene 467 618 46 17 NA #FFFFFF ## 30 AT3G55410 AT3G55410 gene 661 574 46 17 NA #FFFFFF ## 31 AT3G55410 AT3G55410 gene 530 575 46 17 NA #FFFFFF ## 32 AT4G26910 AT4G26910 gene 403 574 46 17 NA #FFFFFF ## 33 AT2G20420 AT2G20420 gene 260 574 46 17 NA #FFFFFF ## 34 AT2G20420 AT2G20420 gene 260 555 46 17 NA #FFFFFF ## 35 AT1G54340 ICDH gene 718 505 46 17 NA #FFFFFF ## 36 AT5G03290 IDH-IV gene 766 458 46 17 -0.4385 #5FDF5F ## 37 AT1G54340 ICDH gene 718 400 46 17 NA #FFFFFF ## 38 AT1G10670 ACLA-3 gene 434 355 46 17 -1.1941 #00FF00 ## 39 AT2G47510 FUM1 gene 191 436 46 17 NA #FFFFFF ## 40 AT2G05710 ACO3 gene 670 345 46 17 NA #FFFFFF ## 41 AT2G05710 ACO3 gene 571 345 46 17 NA #FFFFFF ## 42 AT2G42790 CSY3 gene 434 334 46 17 NA #FFFFFF ## 43 AT1G04410 AT1G04410 gene 253 344 46 17 NA #FFFFFF ## 77 AT1G30120 PDH-E1_ALPHA gene 686 241 46 17 0.6117 #EF3030 ## 79 AT1G30120 PDH-E1_ALPHA gene 582 241 46 17 0.6117 #EF3030 ## 82 AT1G48030 mtLPD1 gene 529 285 46 17 NA #FFFFFF ## 84 AT1G34430 EMB3003 gene 464 242 46 17 NA #FFFFFF ## 91 AT4G37870 PCK1 gene 371 138 46 17 NA #FFFFFF ## 97 AT5G66760 SDH1-2 gene 191 530 46 17 1.1329 #FF0000 ## ## $plot.data.cpd ## kegg.names labels type x y width height FC mol.col ## 58 C00022 C00022 compound 737 241 8 8 0.21082 #CECE8F ## 59 C00122 C00122 compound 190 481 8 8 NA #FFFFFF ## 60 C00036 C00036 compound 320 344 8 8 1.42760 #FFFF00 ## 61 C05379 C05379 compound 718 453 8 8 0.76991 #EFEF30 ## 62 C00024 C00024 compound 400 241 8 8 NA #FFFFFF ## 63 C00149 C00149 compound 190 393 8 8 1.93744 #FFFF00 ## 64 C00311 C00311 compound 718 344 8 8 -1.33400 #0000FF ## 65 C00417 C00417 compound 621 344 8 8 -0.82811 #0000FF ## 66 C00042 C00042 compound 190 574 8 8 1.28575 #FFFF00 ## 67 C00158 C00158 compound 522 344 8 8 NA #FFFFFF ## 68 C15972 C15972 compound 522 617 8 8 NA #FFFFFF ## 69 C00068 C00068 compound 589 530 8 8 0.01167 #BEBEBE ## 70 C16254 C16254 compound 462 573 8 8 NA #FFFFFF ## 71 C15973 C15973 compound 408 617 8 8 NA #FFFFFF ## 72 C00091 C00091 compound 334 574 8 8 -1.97370 #0000FF ## 73 C00026 C00026 compound 718 573 8 8 1.10432 #FFFF00 ## 74 C05381 C05381 compound 593 573 8 8 NA #FFFFFF ## 76 C05125 C05125 compound 631 241 8 8 NA #FFFFFF ## 78 C00068 C00068 compound 631 198 8 8 0.01167 #BEBEBE ## 80 C15972 C15972 compound 580 284 8 8 NA #FFFFFF ## 83 C16255 C16255 compound 517 241 8 8 NA #FFFFFF ## 93 C15973 C15973 compound 478 284 8 8 NA #FFFFFF ## 94 C00074 C00074 compound 471 127 8 8 NA #FFFFFF ``` We can also display changes in genes/proteins and metabolites for multiple comparisons. To do this lets create some more artificial data and this time lets get protein information. ```r metabolite.data2 <- sim.mol.data(mol.type = "cpd", nmol = 3000, nexp = 2) head(metabolite.data2) ``` ``` ## exp1 exp2 ## C02787 -1.15260 1.2914 ## C08521 0.46416 0.6706 ## C01043 0.72893 -0.3065 ## C11496 0.41062 -2.1996 ## C07111 -1.46115 -0.3476 ## C00031 -0.01891 0.7219 ``` ```r gene.data2 <- sim.mol.data(mol.type = "gene", nmol = 3000, nexp = 2, id.type = "UNIPROT") head(gene.data2) ``` ``` ## exp1 exp2 ## E9PDR7 -1.15260 1.2914 ## Q9UBV2 0.46416 0.6706 ## Q9BQ95 0.72893 -0.3065 ## P24310 0.41062 -2.1996 ## Q8N5I0 -1.46115 -0.3476 ## Q9Y6J0 -0.01891 0.7219 ``` Notice UNIPROT IDs do not map to ``arabidopsis thaliana``. We can check the available mappings by looking at the database of identifiers for ``arabidopsis thaliana`` org.At.tair.db. Lets instead map changes to a human pathway or KEGG code "hsa". To do this we need to make sure we supply the correct ``species`` and ``gene.idtype`` arguments. Next lets map changes in proteins and metabolites for the two comparisons to the pathway . ```r map <- gsub("path:ath", "", names(pathways)[1]) # remove 'path:ath' pv.out <- pathview(gene.data = gene.data2, cpd.data = metabolite.data2, gene.idtype = "UNIPROT", pathway.id = map, species = "hsa", out.suffix = map, keys.align = "y", kegg.native = T, match.data = T, key.pos = "topright") plot.name <- paste(map, map, "multi", "png", sep = ".") ``` This should have generated an image file named ``00010.00010.multi.png``. ![hsa00010.00010.multi.png](hsa00010.00010.multi.png) We could also map metabolites on a large scale using `pathway.id= '01100'`. ![metabolic pathways](http://pathview.r-forge.r-project.org/images/hsa01100.sim.cpd.png) ```r sessionInfo() ``` ``` ## R version 3.0.1 (2013-05-16) ## Platform: i386-w64-mingw32/i386 (32-bit) ## ## locale: ## [1] LC_COLLATE=English_United States.1252 ## [2] LC_CTYPE=English_United States.1252 ## [3] LC_MONETARY=English_United States.1252 ## [4] LC_NUMERIC=C ## [5] LC_TIME=English_United States.1252 ## ## attached base packages: ## [1] parallel stats graphics grDevices utils datasets methods ## [8] base ## ## other attached packages: ## [1] KEGGREST_1.2.2 pathview_1.2.4 org.Hs.eg.db_2.9.0 ## [4] RSQLite_0.11.3 DBI_0.2-7 AnnotationDbi_1.24.0 ## [7] Biobase_2.22.0 BiocGenerics_0.8.0 KEGGgraph_1.16.0 ## [10] graph_1.40.1 XML_3.96-1.1 knitr_1.5 ## [13] rcom_2.2-5 rscproxy_2.0-5 ## ## loaded via a namespace (and not attached): ## [1] Biostrings_2.30.1 evaluate_0.5.3 formatR_0.10 ## [4] grid_3.0.1 httr_0.3 IRanges_1.20.7 ## [7] png_0.1-7 RCurl_1.95-4.1 Rgraphviz_2.6.0 ## [10] stats4_3.0.1 stringr_0.6.2 tools_3.0.1 ## [13] XVector_0.2.0 ``` © Dmitry Grapov (2014) Creative Commons License ================================================ FILE: Demos/Pathway Analysis/ath00020.xml ================================================ ================================================ FILE: Demos/Pathway Analysis/ath00260.path ================================================ ================================================ FILE: Demos/Pathway Analysis/ath00260.xml ================================================ ================================================ FILE: Demos/Pathway Analysis/ath00650.xml ================================================ ================================================ FILE: Demos/Pathway Analysis/athp ================================================ ================================================ FILE: Demos/Pathway Analysis/hsa00010.xml ================================================ ================================================ FILE: Demos/Predictive Modeling/Iris O-PLS-DA/O-PLS modeling of Iris data.Rmd ================================================ --- output: html_document: keep_md: yes --- Orthogonal Signal Correction Partial Least Squares (O-PLS) Discriminant Analysis (O-PLS-DA) ======================================================== See here for more [O-PLS](https://raw.github.com/dgrapov/devium/master/R/Devium%20PLS%20%20and%20OPLS.r) methods or try [PLS](http://cran.r-project.org/web/packages/pls/index.html). ```{r,hide=TRUE,message=FALSE,warning=FALSE} source("http://pastebin.com/raw.php?i=UyDBTA57") # source Devium ``` O-PLS-DA demo using Iris data The goal is to predict the species of the flower based on four physical properties. This demonstration is focused on O-PLS-DA model: - [data pretreatment and preparation](#prep) - [model optimization](#oplsda) - [permutation testing](#perm) - [internal cross-validation](#intOOB) - [external cross-validation](#extOOB) - [Comparison of O-PLS-DA and Random Forest classification](#vergin_versi) (updated 11/7/15) ## Preparation for modeling ```{r} data(iris) tmp.data<-iris[,-5] tmp.group<-iris[,5] # species tmp.y<-matrix(as.numeric(tmp.group),ncol=1) # make numeric matrix ``` The data will be split into 1/3 test and 2/3 training sets. The training data will be used for: - model optimization - permutation testing - internally cross-validated estimate of training and out-of-bag error (OOB) The hold out set or the test data will be used to estimate the externally validated OOB. Generate external test set using the duplex or kennard stone method. ```{r} train.test.index.main=test.train.split(nrow(tmp.data),n=1,strata=tmp.group,split.type="duplex",data=tmp.data) train.id<-train.test.index.main=="train" #partition data to get the trainning set tmp.data<-tmp.data[train.id,] tmp.group<-tmp.group[train.id] tmp.y<-tmp.y[train.id,] #the variables could be scaled now, or done internally in the model for each CV split (leave-one-out) #scaled.data<-data.frame(scale(tmp.data,center=TRUE, scale=TRUE)) scaled.data<-tmp.data ``` ## Train O-PLS-DA model Compare a 2 latent variable (LV) PLS-DA and 2 LV with one orthogonal LV (OLV) O-PLS-DA model. ```{r,warning=FALSE,message=FALSE} mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO",method="oscorespls", cv.scale=TRUE, progress=FALSE) #extract model final<-get.OSC.model(obj=mods,OSC.comp=1) #view out-of-bag error for cross-validation splits plot.OSC.results(mods,plot="RMSEP",groups=tmp.group) ``` Ideally we want to select the simplest models with lowest root mean squared error of prediction (RMSEP), which in this case is calculated based on leave-one-one cross-validation. The 1 LV and 1 orthogonal LV (O-PLS-DA) model has similar error to a simple 2 LV PLS-DA model. Next we can also compare the change in scores with the addition of the OLV. ```{r,warning=FALSE, message= FALSE} plot.OSC.results(mods,plot="scores",groups=tmp.group) ``` Non-overlapping scores for each species could signify a well fit model, but we need to carry out some further validations to be sure. ```{r,warning=FALSE, message= FALSE} # plot.PLS.results(obj=final,plot="scores",groups=tmp.group) plot.PLS(obj=final, results = "scores", color=data.frame(species=tmp.group), group.bounds="ellipse",g.alpha=.5, label=FALSE) ``` Ideally the within species variance should be maximally orthogonal to the between species variance (our goal to maximize)in this case we see this represented by the vertical spread of the three species scores. # Permutation Testing Next we can compare our model fit to random chance. We can do this using permutation testing by generating models for a randomly permuted species label (Y). We will generate 50 permuted models and compare their performance statics to our model. We also first want to generate a pseudo-training/test split for our training data in order to correctly simulate the RMSEP. The permuted models will be fit using the pseudo-training data and then used to predict the species label for the pseudo-test set. ```{r,warning=FALSE, message= FALSE} train.test.index<- test.train.split(nrow(scaled.data),n=100,strata=as.factor(tmp.y)) # strata controls if the species are sampled from equally permuted.stats<- permute.OSC.PLS(data=scaled.data,y=as.matrix(tmp.y),n=50,ncomp=2,osc.comp=1, progress=FALSE,train.test.index=train.test.index) #look how our model compares to random chance OSC.validate.model(model=final,perm=permuted.stats) ``` - Q2 represents the in-bag or error for the training data - Xvar the variance in the variables (X) explained or captured in the model - RMSEP is the out-of-bag error (OOB) - The p-values are from a single-sample t-Test comparing our models performance parameters (single values) to their respective permuted distributions # Internal (training set) Cross-validation Next we can estimate the OOB error within the training set by conducting model training and testing. This is done using the pseudo-training/test split we generated for the model permutations. ```{r,warning=FALSE, message= FALSE} train.stats<-OSC.PLS.train.test(pls.data = scaled.data,pls.y = tmp.y,train.test.index ,comp=2,OSC.comp=1,cv.scale=TRUE, progress=FALSE) ``` Now we can compare the distributions for our models' performance statistics to their respective permuted distributions. ```{r,warning=FALSE, message= FALSE} OSC.validate.model(model=mods,perm=permuted.stats,train=train.stats) ``` This suggests that we have a strong model (far better than random chance) which capable of correctly predicting the species of the flower. # Model testing Finally we want to estimate the true estimate of the OOB error by predicting the species labels for the test set we excluded from our data before we even started modeling. The idea is that the test set was never involved in any of our modeling decisions (pretreatment, feature selection, etc) and is the most honest estimate of our models predictive performance. ```{r,warning=FALSE, message= FALSE} #reset data scaled.data<-iris[,-5] tmp.group<-iris[,5] tmp.y<-matrix(as.numeric(tmp.group),ncol=1) #make predictions for the test set mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO", method="oscorespls", cv.scale=TRUE, progress=FALSE,train.test.index=train.test.index.main) #get the true (actual) and predicted values #round them to integers to represent discreet species labels plot.data=data.frame(predicted = round(mods$predicted.Y[[2]][,1],0),actual= mods$test.y) #note these are numeric but we would prefer to interpret classification of species a class plot.data$predicted<-factor(plot.data$predicted,labels=levels(iris[,5]),levels=1:3) plot.data$actual<-factor(plot.data$actual,labels=levels(iris[,5]),levels=1:3) table(plot.data) ``` Based on the similarity between virginica and versicolor species' physical properties we expect this to be the most difficult classification to get correct. ```{r,warning=FALSE, message= FALSE} pairs(iris[,-5],pch=21,bg=rainbow(nlevels(iris[,5]),alpha=.75)[iris[,5]],upper.panel=NULL,cex=2) par(xpd=TRUE) legend(.75,1,levels(iris[,5]),fill=rainbow(nlevels(iris[,5]),alpha=.75),bty="n") ``` O-PLS-DA becomes very useful when there are many variables. Model loadings on LV 1 (x-axis) can be used to linearly rank variables with respect to explaining differences in samples which are represented by the model scores. This approach is useful for reducing multidimensional comparisons (e.g. multiple class comparisons, genotype/treatment/time point) to single dimensional ranking of each variable representing it's weight for explaining a given hypothesis (Y) and the goodness of the answer depends on the strength of the produced model. #### Lets see if we can build a model to discriminate between verginica and versicolor and compare performance between O-PLS-DA and Random Forest (RF). #### Predict classes using O-PLS-DA Lets split the data into train (2/3) test (1/3) sets and compare O-PLS-DA to Random Forest classification performance. ```{r,warning=FALSE, message= FALSE} library(dplyr) library(caret) # set up data data<-iris %>% filter(Species != "setosa") %>% droplevels() #make numeric group<-data$Species %>% as.numeric() #convert to (horrible idea "train", "test") set.seed(106) inTraining <- createDataPartition(group, p = 2/3, list = FALSE) train.index<-rep("test",nrow(data)) train.index[inTraining]<-"train" mods<-make.OSC.PLS.model(group, pls.data=data %>% select(-Species), train.test.index = train.index, comp=2,OSC.comp=1, validation = "LOO", method="oscorespls", cv.scale=TRUE, progress=FALSE) #extract model final<-get.OSC.model(obj=mods,OSC.comp=1) #get performance for test set pred<-mods$predicted.Y[[2]] %>% matrix() %>% #predictions need to be made descreet round(0) #constrained to original bounds pred[pred<0]<-1 pred[pred>2]<-2 obs<-mods$test.y #back convert from numeric obs<-obs %>% factor(.,labels=c("versicolor","verginica")) pred<-pred %>% factor(.,labels=c("versicolor","verginica")) table(pred,obs) ``` #### The performance does not look bad. Lets visualize the misclassified sample using PCA. ```{r,warning=FALSE, message= FALSE} #PCA to visualize test and train samples args<-list( pca.data = data, pca.algorithm = "svd", pca.components = 2, pca.center = FALSE, #need to fix requirement... pca.scaling = "none" ) #calculate and view scree plot res<-devium.pca.calculate(args,return="list",plot=TRUE) #set up plot #show groups group<-data$Species group<-join.columns(data.frame(group,train.index),"_") #id misclassified samples id<-seq_along(1:nrow(data)) id<-id[!id %in% inTraining] #mis miss<-id[obs != pred] tmp<-rep("_",length(group)) tmp[miss]<-"wrong" group<-join.columns(data.frame(group,tmp),"_") %>% gsub("__","",.) p.args<-list( pca = res, results = "scores", color = data.frame(group), label=FALSE, group.bounds ="none", size=5 ) do.call("plot.PCA",p.args) ``` #### Now lets fit and evaluate a random forest model using the `caret` library. ```{r,warning=FALSE, message= FALSE} #create a classification model using random forests #generate training/test set group<-data$Species train.data <- data[ inTraining,] test.data <- data[-inTraining,] train.y <- group[ inTraining] %>% droplevels() test.y <- group[ -inTraining] %>% droplevels() #set cross-validation parameters fitControl <- trainControl( method = "repeatedcv", number = 3, ## repeated ten times repeats = 3, classProbs = TRUE, summaryFunction = twoClassSummary ) #fit model to the training data set.seed(825) fit<- train(train.y ~ ., data = train.data, method = "rf", trControl = fitControl, metric = "ROC", tuneLength = 3 ) #fit to the trainning data print(fit) #predict RF #predict the test set pred<-predict(fit,newdata=test.data) prob<-predict(fit,newdata=test.data,type="prob") obs<-test.y table(pred,obs) #twoClassSummary(data=data.frame(obs,pred,prob),lev=levels(pred)) ``` #### The random forest (RF) model does better than O-PLS, but this is pretty siple data set. From my experience RF is far more robust than O-PLS-DA for classification. It nicely handles multiple classes, is robust to over fitting and unlike O-PLS can be calculated in parallel. ================================================ FILE: Demos/Predictive Modeling/Iris O-PLS-DA/O-PLS modeling of Iris data.html ================================================ Orthogonal Signal Correction Partial Least Squares (O-PLS) Discriminant Analysis (O-PLS-DA)

Orthogonal Signal Correction Partial Least Squares (O-PLS) Discriminant Analysis (O-PLS-DA)

See here for more O-PLS methods or try PLS.

source("http://pastebin.com/raw.php?i=UyDBTA57") # source Devium

O-PLS-DA demo using Iris data The goal is to predict the species of the flower based on four physical properties.

This demonstration is focused on O-PLS-DA model:

Preparation for modeling

data(iris)
tmp.data<-iris[,-5]
tmp.group<-iris[,5] # species
tmp.y<-matrix(as.numeric(tmp.group),ncol=1) # make numeric matrix

The data will be split into 1/3 test and 2/3 training sets. The training data will be used for:

The hold out set or the test data will be used to estimate the externally validated OOB.

Generate external test set using the duplex or kennard stone method.

train.test.index.main=test.train.split(nrow(tmp.data),n=1,strata=tmp.group,split.type="duplex",data=tmp.data)
train.id<-train.test.index.main=="train"

#partition data to get the trainning set
tmp.data<-tmp.data[train.id,]
tmp.group<-tmp.group[train.id]
tmp.y<-tmp.y[train.id,]

#the variables could be scaled now, or done internally in the model for each CV split (leave-one-out)
#scaled.data<-data.frame(scale(tmp.data,center=TRUE, scale=TRUE)) 
scaled.data<-tmp.data

Train O-PLS-DA model

Compare a 2 latent variable (LV) PLS-DA and 2 LV with one orthogonal LV (OLV) O-PLS-DA model.

mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO",method="oscorespls", cv.scale=TRUE, progress=FALSE)
#extract model
final<-get.OSC.model(obj=mods,OSC.comp=1)
#view out-of-bag error for cross-validation splits
plot.OSC.results(mods,plot="RMSEP",groups=tmp.group)

plot of chunk unnamed-chunk-4

Ideally we want to select the simplest models with lowest root mean squared error of prediction (RMSEP), which in this case is calculated based on leave-one-one cross-validation.

The 1 LV and 1 orthogonal LV (O-PLS-DA) model has similar error to a simple 2 LV PLS-DA model.

Next we can also compare the change in scores with the addition of the OLV. 

plot.OSC.results(mods,plot="scores",groups=tmp.group)

plot of chunk unnamed-chunk-5 Non-overlapping scores for each species could signify a well fit model, but we need to carry out some further validations to be sure.

plot.PLS.results(obj=final,plot="scores",groups=tmp.group)

plot of chunk unnamed-chunk-6 Ideally the within species variance should be maximally orthogonal to the between species variance (our goal to maximize)in this case we see this represented by the vertical spread of the three species scores.

Permutation Testing

Next we can compare our model fit to random chance. We can do this using permutation testing by generating models for a randomly permuted species label (Y). We will generate 50 permuted models and compare their performance statics to our model. We also first want to generate a pseudo-training/test split for our training data in order to correctly simulate the RMSEP. The permuted models will be fit using the pseudo-training data and then used to predict the species label for the pseudo-test set.

train.test.index=test.train.split(nrow(scaled.data),n=100,strata=as.factor(tmp.y)) # strata controls if the species are sampled from equally
permuted.stats<-permute.OSC.PLS(data=scaled.data,y=as.matrix(tmp.y),n=50,ncomp=2,osc.comp=1, progress=FALSE,train.test.index=train.test.index)
#look how our model compares to random chance
OSC.validate.model(model=mods,perm=permuted.stats)

## Error: non-numeric argument to binary operator

Internal (training set) Cross-validation

Next we can estimate the OOB error within the training set by conducting model training and testing. This is done using the pseudo-training/test split we generated for the model permutations. 

train.stats<-OSC.PLS.train.test(pls.data = scaled.data,pls.y  = tmp.y,train.test.index ,comp=2,OSC.comp=1,cv.scale=TRUE, progress=FALSE)

Now we can compare the distributions for our models' performance statistics to their respective permuted distributions.

OSC.validate.model(model=mods,perm=permuted.stats,train=train.stats)

##                         Xvar                Q2           RMSEP
## model          99.58 ± 0.178   0.8883 ± 0.0537 0.2342 ± 0.0516
## permuted model 99.29 ± 0.457 -0.06332 ± 0.0444 0.8438 ± 0.0566
## p-value            8.437e-05        1.183e-121       3.731e-77

This suggests that we have a strong model (far better than random chance) which capable of correctly predicting the species of the flower.

Model testing

Finally we want to estimate the true estimate of the OOB error by predicting the species labels for the test set we excluded from our data before we even started modeling.

The idea is that the test set was never involved in any of our modeling decisions (pretreatment, feature selection, etc) and is the most honest estimate of our models predictive performance.

#reset data
scaled.data<-iris[,-5]
tmp.group<-iris[,5] 
tmp.y<-matrix(as.numeric(tmp.group),ncol=1)  

#make predictions for the test set
mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO",
method="oscorespls", cv.scale=TRUE, progress=FALSE,train.test.index=train.test.index.main)

#get the true (actual) and predicted values
#round them to integers to represent discreet species labels
plot.data=data.frame(predicted = round(mods$predicted.Y[[2]][,1],0),actual= mods$test.y)
#note these are numeric but we would prefer to interpret classification of species a class
plot.data$predicted<-factor(plot.data$predicted,labels=levels(iris[,5]),levels=1:3)
plot.data$actual<-factor(plot.data$actual,labels=levels(iris[,5]),levels=1:3)

table(plot.data)

##             actual
## predicted    setosa versicolor virginica
##   setosa         17          0         0
##   versicolor      0         16         1
##   virginica       0          1        16

Based on the similarity between virginica and versicolor species' physical properties we expect this to be the most difficult classification to get correct.

pairs(iris[,-5],pch=21,bg=rainbow(nlevels(iris[,5]),alpha=.75)[iris[,5]],upper.panel=NULL,cex=2)
par(xpd=TRUE)
legend(.75,1,levels(iris[,5]),fill=rainbow(nlevels(iris[,5]),alpha=.75),bty="n")

plot of chunk unnamed-chunk-11

O-PLS-DA becomes very useful when there are many variables. Model loadings on LV 1 (x-axis) can be used to linearly rank variables with respect to explaining differences in samples which are represented by the model scores. This approach is useful for reducing multidimensional comparisons (e.g. multiple class comparisons, genotype/treatment/time point) to single dimensional ranking of each variable representing it's weight for explaining a given hypothesis (Y) and the goodness of the answer depends on the strength of the produced model.

================================================ FILE: Demos/Predictive Modeling/Iris O-PLS-DA/O-PLS modeling of Iris data.md ================================================ Orthogonal Signal Correction Partial Least Squares (O-PLS) Discriminant Analysis (O-PLS-DA) ======================================================== See here for more [O-PLS](https://raw.github.com/dgrapov/devium/master/R/Devium%20PLS%20%20and%20OPLS.r) methods or try [PLS](http://cran.r-project.org/web/packages/pls/index.html). ```r source("http://pastebin.com/raw.php?i=UyDBTA57") # source Devium ``` O-PLS-DA demo using Iris data The goal is to predict the species of the flower based on four physical properties. This demonstration is focused on O-PLS-DA model: - [data pretreatment and preparation](#prep) - [model optimization](#oplsda) - [permutation testing](#perm) - [internal cross-validation](#intOOB) - [external cross-validation](#extOOB) ## Preparation for modeling ```r data(iris) tmp.data<-iris[,-5] tmp.group<-iris[,5] # species tmp.y<-matrix(as.numeric(tmp.group),ncol=1) # make numeric matrix ``` The data will be split into 1/3 test and 2/3 training sets. The training data will be used for: - model optimization - permutation testing - internally cross-validated estimate of training and out-of-bag error (OOB) The hold out set or the test data will be used to estimate the externally validated OOB. Generate external test set using the duplex or kennard stone method. ```r train.test.index.main=test.train.split(nrow(tmp.data),n=1,strata=tmp.group,split.type="duplex",data=tmp.data) train.id<-train.test.index.main=="train" #partition data to get the trainning set tmp.data<-tmp.data[train.id,] tmp.group<-tmp.group[train.id] tmp.y<-tmp.y[train.id,] #the variables could be scaled now, or done internally in the model for each CV split (leave-one-out) #scaled.data<-data.frame(scale(tmp.data,center=TRUE, scale=TRUE)) scaled.data<-tmp.data ``` ## Train O-PLS-DA model Compare a 2 latent variable (LV) PLS-DA and 2 LV with one orthogonal LV (OLV) O-PLS-DA model. ```r mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO",method="oscorespls", cv.scale=TRUE, progress=FALSE) #extract model final<-get.OSC.model(obj=mods,OSC.comp=1) #view out-of-bag error for cross-validation splits plot.OSC.results(mods,plot="RMSEP",groups=tmp.group) ``` ![plot of chunk unnamed-chunk-4](figure/unnamed-chunk-4.png) Ideally we want to select the simplest models with lowest root mean squared error of prediction (RMSEP), which in this case is calculated based on leave-one-one cross-validation. The 1 LV and 1 orthogonal LV (O-PLS-DA) model has similar error to a simple 2 LV PLS-DA model. Next we can also compare the change in scores with the addition of the OLV. ```r plot.OSC.results(mods,plot="scores",groups=tmp.group) ``` ![plot of chunk unnamed-chunk-5](figure/unnamed-chunk-5.png) Non-overlapping scores for each species could signify a well fit model, but we need to carry out some further validations to be sure. ```r plot.PLS.results(obj=final,plot="scores",groups=tmp.group) ``` ![plot of chunk unnamed-chunk-6](figure/unnamed-chunk-6.png) Ideally the within species variance should be maximally orthogonal to the between species variance (our goal to maximize)in this case we see this represented by the vertical spread of the three species scores. # Permutation Testing Next we can compare our model fit to random chance. We can do this using permutation testing by generating models for a randomly permuted species label (Y). We will generate 50 permuted models and compare their performance statics to our model. We also first want to generate a pseudo-training/test split for our training data in order to correctly simulate the RMSEP. The permuted models will be fit using the pseudo-training data and then used to predict the species label for the pseudo-test set. ```r train.test.index=test.train.split(nrow(scaled.data),n=100,strata=as.factor(tmp.y)) # strata controls if the species are sampled from equally permuted.stats<-permute.OSC.PLS(data=scaled.data,y=as.matrix(tmp.y),n=50,ncomp=2,osc.comp=1, progress=FALSE,train.test.index=train.test.index) #look how our model compares to random chance OSC.validate.model(model=mods,perm=permuted.stats) ``` ``` ## Error: non-numeric argument to binary operator ``` - Q2 represents the in-bag or error for the training data - Xvar the variance in the variables (X) explained or captured in the model - RMSEP is the out-of-bag error (OOB) - The p-values are from a single-sample t-Test comparing our models performance parameters (single values) to their respective permuted distributions # Internal (training set) Cross-validation Next we can estimate the OOB error within the training set by conducting model training and testing. This is done using the pseudo-training/test split we generated for the model permutations. ```r train.stats<-OSC.PLS.train.test(pls.data = scaled.data,pls.y = tmp.y,train.test.index ,comp=2,OSC.comp=1,cv.scale=TRUE, progress=FALSE) ``` Now we can compare the distributions for our models' performance statistics to their respective permuted distributions. ```r OSC.validate.model(model=mods,perm=permuted.stats,train=train.stats) ``` ``` ## Xvar Q2 RMSEP ## model 99.58 ± 0.178 0.8883 ± 0.0537 0.2342 ± 0.0516 ## permuted model 99.29 ± 0.457 -0.06332 ± 0.0444 0.8438 ± 0.0566 ## p-value 8.437e-05 1.183e-121 3.731e-77 ``` This suggests that we have a strong model (far better than random chance) which capable of correctly predicting the species of the flower. # Model testing Finally we want to estimate the true estimate of the OOB error by predicting the species labels for the test set we excluded from our data before we even started modeling. The idea is that the test set was never involved in any of our modeling decisions (pretreatment, feature selection, etc) and is the most honest estimate of our models predictive performance. ```r #reset data scaled.data<-iris[,-5] tmp.group<-iris[,5] tmp.y<-matrix(as.numeric(tmp.group),ncol=1) #make predictions for the test set mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO", method="oscorespls", cv.scale=TRUE, progress=FALSE,train.test.index=train.test.index.main) #get the true (actual) and predicted values #round them to integers to represent discreet species labels plot.data=data.frame(predicted = round(mods$predicted.Y[[2]][,1],0),actual= mods$test.y) #note these are numeric but we would prefer to interpret classification of species a class plot.data$predicted<-factor(plot.data$predicted,labels=levels(iris[,5]),levels=1:3) plot.data$actual<-factor(plot.data$actual,labels=levels(iris[,5]),levels=1:3) table(plot.data) ``` ``` ## actual ## predicted setosa versicolor virginica ## setosa 17 0 0 ## versicolor 0 16 1 ## virginica 0 1 16 ``` Based on the similarity between virginica and versicolor species' physical properties we expect this to be the most difficult classification to get correct. ```r pairs(iris[,-5],pch=21,bg=rainbow(nlevels(iris[,5]),alpha=.75)[iris[,5]],upper.panel=NULL,cex=2) par(xpd=TRUE) legend(.75,1,levels(iris[,5]),fill=rainbow(nlevels(iris[,5]),alpha=.75),bty="n") ``` ![plot of chunk unnamed-chunk-11](figure/unnamed-chunk-11.png) O-PLS-DA becomes very useful when there are many variables. Model loadings on LV 1 (x-axis) can be used to linearly rank variables with respect to explaining differences in samples which are represented by the model scores. This approach is useful for reducing multidimensional comparisons (e.g. multiple class comparisons, genotype/treatment/time point) to single dimensional ranking of each variable representing it's weight for explaining a given hypothesis (Y) and the goodness of the answer depends on the strength of the produced model. ================================================ FILE: Demos/Predictive Modeling/Iris O-PLS-DA/O-PLS_modeling_of_Iris_data.html ================================================

Orthogonal Signal Correction Partial Least Squares (O-PLS) Discriminant Analysis (O-PLS-DA)

See here for more O-PLS methods or try PLS.

source("http://pastebin.com/raw.php?i=UyDBTA57") # source Devium

O-PLS-DA demo using Iris data The goal is to predict the species of the flower based on four physical properties.

This demonstration is focused on O-PLS-DA model: - data pretreatment and preparation - model optimization - permutation testing - internal cross-validation - external cross-validation - Comparison of O-PLS-DA and Random Forest classification (updated 11/7/15)

## Preparation for modeling

data(iris)
tmp.data<-iris[,-5]
tmp.group<-iris[,5] # species
tmp.y<-matrix(as.numeric(tmp.group),ncol=1) # make numeric matrix

The data will be split into 1/3 test and 2/3 training sets. The training data will be used for:

The hold out set or the test data will be used to estimate the externally validated OOB.

Generate external test set using the duplex or kennard stone method.

train.test.index.main=test.train.split(nrow(tmp.data),n=1,strata=tmp.group,split.type="duplex",data=tmp.data)
train.id<-train.test.index.main=="train"

#partition data to get the trainning set
tmp.data<-tmp.data[train.id,]
tmp.group<-tmp.group[train.id]
tmp.y<-tmp.y[train.id,]

#the variables could be scaled now, or done internally in the model for each CV split (leave-one-out)
#scaled.data<-data.frame(scale(tmp.data,center=TRUE, scale=TRUE)) 
scaled.data<-tmp.data

## Train O-PLS-DA model Compare a 2 latent variable (LV) PLS-DA and 2 LV with one orthogonal LV (OLV) O-PLS-DA model.

mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO",method="oscorespls", cv.scale=TRUE, progress=FALSE)
#extract model
final<-get.OSC.model(obj=mods,OSC.comp=1)
#view out-of-bag error for cross-validation splits
plot.OSC.results(mods,plot="RMSEP",groups=tmp.group)

Ideally we want to select the simplest models with lowest root mean squared error of prediction (RMSEP), which in this case is calculated based on leave-one-one cross-validation.

The 1 LV and 1 orthogonal LV (O-PLS-DA) model has similar error to a simple 2 LV PLS-DA model.

Next we can also compare the change in scores with the addition of the OLV.

plot.OSC.results(mods,plot="scores",groups=tmp.group)

Non-overlapping scores for each species could signify a well fit model, but we need to carry out some further validations to be sure.

# plot.PLS.results(obj=final,plot="scores",groups=tmp.group)
plot.PLS(obj=final, results = "scores", color=data.frame(species=tmp.group), group.bounds="ellipse",g.alpha=.5, label=FALSE)

Ideally the within species variance should be maximally orthogonal to the between species variance (our goal to maximize)in this case we see this represented by the vertical spread of the three species scores.

# Permutation Testing Next we can compare our model fit to random chance. We can do this using permutation testing by generating models for a randomly permuted species label (Y). We will generate 50 permuted models and compare their performance statics to our model. We also first want to generate a pseudo-training/test split for our training data in order to correctly simulate the RMSEP. The permuted models will be fit using the pseudo-training data and then used to predict the species label for the pseudo-test set.

train.test.index<- test.train.split(nrow(scaled.data),n=100,strata=as.factor(tmp.y)) # strata controls if the species are sampled from equally
permuted.stats<- permute.OSC.PLS(data=scaled.data,y=as.matrix(tmp.y),n=50,ncomp=2,osc.comp=1, progress=FALSE,train.test.index=train.test.index)
#look how our model compares to random chance
OSC.validate.model(model=final,perm=permuted.stats)
##                            RX2                  Q2             RMSEP
## model                    99.54              0.9226            0.2271
## permuted model 99.41 +/- 0.455 -0.06364 +/- 0.0361 0.8436 +/- 0.0468
## p-value                0.05058           2.664e-72         8.228e-57

# Internal (training set) Cross-validation Next we can estimate the OOB error within the training set by conducting model training and testing. This is done using the pseudo-training/test split we generated for the model permutations.

train.stats<-OSC.PLS.train.test(pls.data = scaled.data,pls.y  = tmp.y,train.test.index ,comp=2,OSC.comp=1,cv.scale=TRUE, progress=FALSE)

Now we can compare the distributions for our models’ performance statistics to their respective permuted distributions.

OSC.validate.model(model=mods,perm=permuted.stats,train=train.stats)
##                           Xvar                  Q2             RMSEP
## model           99.59 +/- 0.35   0.9007 +/- 0.0436 0.2248 +/- 0.0209
## permuted model 99.41 +/- 0.455 -0.06364 +/- 0.0361 0.8436 +/- 0.0468
## p-value                0.01923          1.184e-132         1.331e-64

This suggests that we have a strong model (far better than random chance) which capable of correctly predicting the species of the flower. # Model testing Finally we want to estimate the true estimate of the OOB error by predicting the species labels for the test set we excluded from our data before we even started modeling.

The idea is that the test set was never involved in any of our modeling decisions (pretreatment, feature selection, etc) and is the most honest estimate of our models predictive performance.

#reset data
scaled.data<-iris[,-5]
tmp.group<-iris[,5] 
tmp.y<-matrix(as.numeric(tmp.group),ncol=1)  

#make predictions for the test set
mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO",
method="oscorespls", cv.scale=TRUE, progress=FALSE,train.test.index=train.test.index.main)

#get the true (actual) and predicted values
#round them to integers to represent discreet species labels
plot.data=data.frame(predicted = round(mods$predicted.Y[[2]][,1],0),actual= mods$test.y)
#note these are numeric but we would prefer to interpret classification of species a class
plot.data$predicted<-factor(plot.data$predicted,labels=levels(iris[,5]),levels=1:3)
plot.data$actual<-factor(plot.data$actual,labels=levels(iris[,5]),levels=1:3)

table(plot.data)
##             actual
## predicted    setosa versicolor virginica
##   setosa         17          0         0
##   versicolor      0         15         0
##   virginica       0          2        17

Based on the similarity between virginica and versicolor species’ physical properties we expect this to be the most difficult classification to get correct.

pairs(iris[,-5],pch=21,bg=rainbow(nlevels(iris[,5]),alpha=.75)[iris[,5]],upper.panel=NULL,cex=2)
par(xpd=TRUE)
legend(.75,1,levels(iris[,5]),fill=rainbow(nlevels(iris[,5]),alpha=.75),bty="n")

O-PLS-DA becomes very useful when there are many variables. Model loadings on LV 1 (x-axis) can be used to linearly rank variables with respect to explaining differences in samples which are represented by the model scores. This approach is useful for reducing multidimensional comparisons (e.g. multiple class comparisons, genotype/treatment/time point) to single dimensional ranking of each variable representing it’s weight for explaining a given hypothesis (Y) and the goodness of the answer depends on the strength of the produced model.

#### Lets see if we can build a model to discriminate between verginica and versicolor and compare performance between O-PLS-DA and Random Forest (RF). #### Predict classes using O-PLS-DA Lets split the data into train (2/3) test (1/3) sets and compare O-PLS-DA to Random Forest classification performance.

library(dplyr)
library(caret)
# set up data 
data<-iris %>% filter(Species != "setosa") %>% droplevels()
#make numeric
group<-data$Species %>% as.numeric()

#convert to (horrible idea "train", "test")
set.seed(106)
inTraining <- createDataPartition(group, p = 2/3, list = FALSE)
train.index<-rep("test",nrow(data)) 
train.index[inTraining]<-"train"

mods<-make.OSC.PLS.model(group,
                         pls.data=data %>% select(-Species),
                         train.test.index = train.index,
                         comp=2,OSC.comp=1, validation = "LOO",
                         method="oscorespls", cv.scale=TRUE, progress=FALSE)
#extract model
final<-get.OSC.model(obj=mods,OSC.comp=1)

#get performance for test set
pred<-mods$predicted.Y[[2]] %>% matrix() %>%
#predictions need to be made descreet
  round(0)
#constrained to original bounds
pred[pred<0]<-1
pred[pred>2]<-2
obs<-mods$test.y
#back convert from numeric
obs<-obs %>% factor(.,labels=c("versicolor","verginica"))
pred<-pred %>% factor(.,labels=c("versicolor","verginica"))

table(pred,obs)
##             obs
## pred         versicolor verginica
##   versicolor         15         0
##   verginica           1        16

The performance does not look bad. Lets visualize the misclassified sample using PCA.

#PCA to visualize test and train samples
args<-list(  pca.data       = data,
            pca.algorithm   = "svd",
            pca.components  = 2,
            pca.center      = FALSE, #need to fix requirement...
            pca.scaling     = "none"
            )

#calculate and view scree plot          
res<-devium.pca.calculate(args,return="list",plot=TRUE)

#set up plot
#show groups
group<-data$Species 
group<-join.columns(data.frame(group,train.index),"_")
#id misclassified samples
id<-seq_along(1:nrow(data))
id<-id[!id %in% inTraining]
#mis
miss<-id[obs != pred]
tmp<-rep("_",length(group))
tmp[miss]<-"wrong"
group<-join.columns(data.frame(group,tmp),"_") %>% gsub("__","",.)

p.args<-list( 
            pca = res,
            results = "scores",
            color = data.frame(group),
            label=FALSE,
            group.bounds ="none",
            size=5
            )

do.call("plot.PCA",p.args)

Now lets fit and evaluate a random forest model using the caret library.

#create a classification model using random forests
#generate training/test set
group<-data$Species

train.data <- data[ inTraining,]
test.data  <- data[-inTraining,]
train.y <- group[ inTraining] %>% droplevels()
test.y <- group[ -inTraining] %>% droplevels()

#set cross-validation parameters
fitControl <- trainControl(
                          method = "repeatedcv",
                          number = 3,
                          ## repeated ten times
                          repeats = 3,
                          classProbs = TRUE,
                          summaryFunction = twoClassSummary
                )


#fit model to the training data
set.seed(825)
fit<- train(train.y ~ ., data = train.data,
            method = "rf",
            trControl = fitControl,
            metric = "ROC",
            tuneLength = 3  
         )

#fit to the trainning data
print(fit)
## Random Forest 
## 
## 68 samples
##  4 predictor
##  2 classes: 'versicolor', 'virginica' 
## 
## No pre-processing
## Resampling: Cross-Validated (3 fold, repeated 3 times) 
## 
## Summary of sample sizes: 46, 46, 44, 45, 46, 45, ... 
## 
## Resampling results across tuning parameters:
## 
##   mtry  ROC  Sens  Spec  ROC SD  Sens SD  Spec SD
##   2     1    1     1     0       0        0      
##   3     1    1     1     0       0        0      
##   5     1    1     1     0       0        0      
## 
## ROC was used to select the optimal model using  the largest value.
## The final value used for the model was mtry = 2.
#predict RF
#predict the test set
pred<-predict(fit,newdata=test.data)
prob<-predict(fit,newdata=test.data,type="prob")
obs<-test.y
table(pred,obs)
##             obs
## pred         versicolor virginica
##   versicolor         16         0
##   virginica           0        16
#twoClassSummary(data=data.frame(obs,pred,prob),lev=levels(pred))

The random forest (RF) model does better than O-PLS, but this is pretty siple data set. From my experience RF is far more robust than O-PLS-DA for classification. It nicely handles multiple classes, is robust to over fitting and unlike O-PLS can be calculated in parallel.

================================================ FILE: Demos/Predictive Modeling/Iris O-PLS-DA/O-PLS_modeling_of_Iris_data.md ================================================ Orthogonal Signal Correction Partial Least Squares (O-PLS) Discriminant Analysis (O-PLS-DA) ======================================================== ##### load devium functions from: https://github.com/dgrapov/devium/tree/master/R TODO make a package...for now: ```r source("http://pastebin.com/raw.php?i=UyDBTA57") # source Devium ``` O-PLS-DA demo using Iris data The goal is to predict the species of the flower based on four physical properties. This demonstration is focused on O-PLS-DA model: - [data pretreatment and preparation](#prep) - [model optimization](#oplsda) - [permutation testing](#perm) - [internal cross-validation](#intOOB) - [external cross-validation](#extOOB) - [Comparison of O-PLS-DA and Random Forest classification](#vergin_versi) (updated 11/7/15) ## Preparation for modeling ```r data(iris) tmp.data<-iris[,-5] tmp.group<-iris[,5] # species tmp.y<-matrix(as.numeric(tmp.group),ncol=1) # make numeric matrix ``` The data will be split into 1/3 test and 2/3 training sets. The training data will be used for: - model optimization - permutation testing - internally cross-validated estimate of training and out-of-bag error (OOB) The hold out set or the test data will be used to estimate the externally validated OOB. Generate external test set using the duplex or kennard stone method. ```r train.test.index.main=test.train.split(nrow(tmp.data),n=1,strata=tmp.group,split.type="duplex",data=tmp.data) train.id<-train.test.index.main=="train" #partition data to get the trainning set tmp.data<-tmp.data[train.id,] tmp.group<-tmp.group[train.id] tmp.y<-tmp.y[train.id,] #the variables could be scaled now, or done internally in the model for each CV split (leave-one-out) #scaled.data<-data.frame(scale(tmp.data,center=TRUE, scale=TRUE)) scaled.data<-tmp.data ``` ## Train O-PLS-DA model Compare a 2 latent variable (LV) PLS-DA and 2 LV with one orthogonal LV (OLV) O-PLS-DA model. ```r mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO",method="oscorespls", cv.scale=TRUE, progress=FALSE) #extract model final<-get.OSC.model(obj=mods,OSC.comp=1) #view out-of-bag error for cross-validation splits plot.OSC.results(mods,plot="RMSEP",groups=tmp.group) ``` ![](O-PLS_modeling_of_Iris_data_files/figure-html/unnamed-chunk-4-1.png) Ideally we want to select the simplest models with lowest root mean squared error of prediction (RMSEP), which in this case is calculated based on leave-one-one cross-validation. The 1 LV and 1 orthogonal LV (O-PLS-DA) model has similar error to a simple 2 LV PLS-DA model. Next we can also compare the change in scores with the addition of the OLV. ```r plot.OSC.results(mods,plot="scores",groups=tmp.group) ``` ![](O-PLS_modeling_of_Iris_data_files/figure-html/unnamed-chunk-5-1.png) Non-overlapping scores for each species could signify a well fit model, but we need to carry out some further validations to be sure. ```r # plot.PLS.results(obj=final,plot="scores",groups=tmp.group) plot.PLS(obj=final, results = "scores", color=data.frame(species=tmp.group), group.bounds="ellipse",g.alpha=.5, label=FALSE) ``` ![](O-PLS_modeling_of_Iris_data_files/figure-html/unnamed-chunk-6-1.png) Ideally the within species variance should be maximally orthogonal to the between species variance (our goal to maximize)in this case we see this represented by the vertical spread of the three species scores. # Permutation Testing Next we can compare our model fit to random chance. We can do this using permutation testing by generating models for a randomly permuted species label (Y). We will generate 50 permuted models and compare their performance statics to our model. We also first want to generate a pseudo-training/test split for our training data in order to correctly simulate the RMSEP. The permuted models will be fit using the pseudo-training data and then used to predict the species label for the pseudo-test set. ```r train.test.index<- test.train.split(nrow(scaled.data),n=100,strata=as.factor(tmp.y)) # strata controls if the species are sampled from equally permuted.stats<- permute.OSC.PLS(data=scaled.data,y=as.matrix(tmp.y),n=50,ncomp=2,osc.comp=1, progress=FALSE,train.test.index=train.test.index) #look how our model compares to random chance OSC.validate.model(model=final,perm=permuted.stats) ``` ``` ## RX2 Q2 RMSEP ## model 99.54 0.9226 0.2271 ## permuted model 99.41 +/- 0.455 -0.06364 +/- 0.0361 0.8436 +/- 0.0468 ## p-value 0.05058 2.664e-72 8.228e-57 ``` - Q2 represents the in-bag or error for the training data - Xvar the variance in the variables (X) explained or captured in the model - RMSEP is the out-of-bag error (OOB) - The p-values are from a single-sample t-Test comparing our models performance parameters (single values) to their respective permuted distributions # Internal (training set) Cross-validation Next we can estimate the OOB error within the training set by conducting model training and testing. This is done using the pseudo-training/test split we generated for the model permutations. ```r train.stats<-OSC.PLS.train.test(pls.data = scaled.data,pls.y = tmp.y,train.test.index ,comp=2,OSC.comp=1,cv.scale=TRUE, progress=FALSE) ``` Now we can compare the distributions for our models' performance statistics to their respective permuted distributions. ```r OSC.validate.model(model=mods,perm=permuted.stats,train=train.stats) ``` ``` ## Xvar Q2 RMSEP ## model 99.59 +/- 0.35 0.9007 +/- 0.0436 0.2248 +/- 0.0209 ## permuted model 99.41 +/- 0.455 -0.06364 +/- 0.0361 0.8436 +/- 0.0468 ## p-value 0.01923 1.184e-132 1.331e-64 ``` This suggests that we have a strong model (far better than random chance) which capable of correctly predicting the species of the flower. # Model testing Finally we want to estimate the true estimate of the OOB error by predicting the species labels for the test set we excluded from our data before we even started modeling. The idea is that the test set was never involved in any of our modeling decisions (pretreatment, feature selection, etc) and is the most honest estimate of our models predictive performance. ```r #reset data scaled.data<-iris[,-5] tmp.group<-iris[,5] tmp.y<-matrix(as.numeric(tmp.group),ncol=1) #make predictions for the test set mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO", method="oscorespls", cv.scale=TRUE, progress=FALSE,train.test.index=train.test.index.main) #get the true (actual) and predicted values #round them to integers to represent discreet species labels plot.data=data.frame(predicted = round(mods$predicted.Y[[2]][,1],0),actual= mods$test.y) #note these are numeric but we would prefer to interpret classification of species a class plot.data$predicted<-factor(plot.data$predicted,labels=levels(iris[,5]),levels=1:3) plot.data$actual<-factor(plot.data$actual,labels=levels(iris[,5]),levels=1:3) table(plot.data) ``` ``` ## actual ## predicted setosa versicolor virginica ## setosa 17 0 0 ## versicolor 0 15 0 ## virginica 0 2 17 ``` Based on the similarity between virginica and versicolor species' physical properties we expect this to be the most difficult classification to get correct. ```r pairs(iris[,-5],pch=21,bg=rainbow(nlevels(iris[,5]),alpha=.75)[iris[,5]],upper.panel=NULL,cex=2) par(xpd=TRUE) legend(.75,1,levels(iris[,5]),fill=rainbow(nlevels(iris[,5]),alpha=.75),bty="n") ``` ![](O-PLS_modeling_of_Iris_data_files/figure-html/unnamed-chunk-11-1.png) O-PLS-DA becomes very useful when there are many variables. Model loadings on LV 1 (x-axis) can be used to linearly rank variables with respect to explaining differences in samples which are represented by the model scores. This approach is useful for reducing multidimensional comparisons (e.g. multiple class comparisons, genotype/treatment/time point) to single dimensional ranking of each variable representing it's weight for explaining a given hypothesis (Y) and the goodness of the answer depends on the strength of the produced model. #### Lets see if we can build a model to discriminate between verginica and versicolor and compare performance between O-PLS-DA and Random Forest (RF). #### Predict classes using O-PLS-DA Lets split the data into train (2/3) test (1/3) sets and compare O-PLS-DA to Random Forest classification performance. ```r library(dplyr) library(caret) # set up data data<-iris %>% filter(Species != "setosa") %>% droplevels() #make numeric group<-data$Species %>% as.numeric() #convert to (horrible idea "train", "test") set.seed(106) inTraining <- createDataPartition(group, p = 2/3, list = FALSE) train.index<-rep("test",nrow(data)) train.index[inTraining]<-"train" mods<-make.OSC.PLS.model(group, pls.data=data %>% select(-Species), train.test.index = train.index, comp=2,OSC.comp=1, validation = "LOO", method="oscorespls", cv.scale=TRUE, progress=FALSE) #extract model final<-get.OSC.model(obj=mods,OSC.comp=1) #get performance for test set pred<-mods$predicted.Y[[2]] %>% matrix() %>% #predictions need to be made descreet round(0) #constrained to original bounds pred[pred<0]<-1 pred[pred>2]<-2 obs<-mods$test.y #back convert from numeric obs<-obs %>% factor(.,labels=c("versicolor","verginica")) pred<-pred %>% factor(.,labels=c("versicolor","verginica")) table(pred,obs) ``` ``` ## obs ## pred versicolor verginica ## versicolor 15 0 ## verginica 1 16 ``` #### The performance does not look bad. Lets visualize the misclassified sample using PCA. ```r #PCA to visualize test and train samples args<-list( pca.data = data, pca.algorithm = "svd", pca.components = 2, pca.center = FALSE, #need to fix requirement... pca.scaling = "none" ) #calculate and view scree plot res<-devium.pca.calculate(args,return="list",plot=TRUE) ``` ![](O-PLS_modeling_of_Iris_data_files/figure-html/unnamed-chunk-13-1.png) ```r #set up plot #show groups group<-data$Species group<-join.columns(data.frame(group,train.index),"_") #id misclassified samples id<-seq_along(1:nrow(data)) id<-id[!id %in% inTraining] #mis miss<-id[obs != pred] tmp<-rep("_",length(group)) tmp[miss]<-"wrong" group<-join.columns(data.frame(group,tmp),"_") %>% gsub("__","",.) p.args<-list( pca = res, results = "scores", color = data.frame(group), label=FALSE, group.bounds ="none", size=5 ) do.call("plot.PCA",p.args) ``` ![](O-PLS_modeling_of_Iris_data_files/figure-html/unnamed-chunk-13-2.png) #### Now lets fit and evaluate a random forest model using the `caret` library. ```r #create a classification model using random forests #generate training/test set group<-data$Species train.data <- data[ inTraining,] test.data <- data[-inTraining,] train.y <- group[ inTraining] %>% droplevels() test.y <- group[ -inTraining] %>% droplevels() #set cross-validation parameters fitControl <- trainControl( method = "repeatedcv", number = 3, ## repeated ten times repeats = 3, classProbs = TRUE, summaryFunction = twoClassSummary ) #fit model to the training data set.seed(825) fit<- train(train.y ~ ., data = train.data, method = "rf", trControl = fitControl, metric = "ROC", tuneLength = 3 ) #fit to the trainning data print(fit) ``` ``` ## Random Forest ## ## 68 samples ## 4 predictor ## 2 classes: 'versicolor', 'virginica' ## ## No pre-processing ## Resampling: Cross-Validated (3 fold, repeated 3 times) ## ## Summary of sample sizes: 46, 46, 44, 45, 46, 45, ... ## ## Resampling results across tuning parameters: ## ## mtry ROC Sens Spec ROC SD Sens SD Spec SD ## 2 1 1 1 0 0 0 ## 3 1 1 1 0 0 0 ## 5 1 1 1 0 0 0 ## ## ROC was used to select the optimal model using the largest value. ## The final value used for the model was mtry = 2. ``` ```r #predict RF #predict the test set pred<-predict(fit,newdata=test.data) prob<-predict(fit,newdata=test.data,type="prob") obs<-test.y table(pred,obs) ``` ``` ## obs ## pred versicolor virginica ## versicolor 16 0 ## virginica 0 16 ``` ```r #twoClassSummary(data=data.frame(obs,pred,prob),lev=levels(pred)) ``` #### The random forest (RF) model does better than O-PLS, but this is a pretty simple data set. From my experience RF is far more robust than O-PLS-DA for classification. It nicely handles multiple classes, is robust to over fitting and unlike O-PLS can be calculated in parallel. ================================================ FILE: Demos/Predictive Modeling/Iris O-PLS-DA/demo.r ================================================ # Orthogonal Signal Correction Partial Least Squares Discriminant Analysis # O-PLS-DA demo using Iris data source("http://pastebin.com/raw.php?i=JVyTrYRD") # load functions # The goal is to predict the species of the flower based on four physical properties. data(iris) tmp.data<-iris[,-5] tmp.group<-iris[,5] # species tmp.y<-matrix(as.numeric(tmp.group),ncol=1) # make numeric matrix # The data will be split into 1/3 test and 2/3 trainning sets # Model performance will be estimated based # generate external test set train.test.index.main=test.train.split(nrow(tmp.data),n=1,strata=tmp.group,split.type="duplex",data=tmp.data) train.id<-train.test.index.main=="train" #partition data to get the trainning set tmp.data<-tmp.data[train.id,] tmp.group<-tmp.group[train.id] tmp.y<-tmp.y[train.id,] #the variables could be scaled now, or done internally in the model for each CV split (leave-one-out) # scaled.data<-data.frame(scale(tmp.data,center=TRUE, scale=TRUE)) scaled.data<-tmp.data #make OSC model mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=3,OSC.comp=2, validation = "LOO",method="oscorespls", cv.scale=TRUE, progress=FALSE) # view out-of-bag error for cross-validation splits plot.OSC.results(mods,plot="RMSEP",groups=tmp.group) # the red line shows the score for a 3 component or latent variable (LV) PLS model with no orthogonal components # because the dependant variable (Y) is discreet this is simply called PLS-DA # looking for the simplest models with lowest error we could choose # the 2 LV and 1 orthogonal LV (OLV) model, which has the lowest root mean squared error of prediction (RMSEP) mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO",method="oscorespls", cv.scale=TRUE, progress=FALSE) # #we can also look at the change in model scores for each sample as we increase the number of OLVs plot.OSC.results(mods,plot="scores",groups=tmp.group) # The goal is to capture the maximum separation between groups in the x-axis or the predictive LV # Species scores for the chosen model look well resolved. # This suggest that we are off to a good start # or more specifically that the variables we have can be used to classify species of Iris flowers. # Ideally the within species variance should be maximally orthogonal to the between species variance (goal to maximize) # in this case we see this represented by the vertical spread of the three species scores #The chosen model can be extracted and reploted to show individual sample scores final<-results<-get.OSC.model(obj=mods,OSC.comp=1) plot.PLS.results(obj=final,plot="scores",groups=tmp.group) # next we can check how likely is our fit by generating # model statistics while randomly shuffling the Y (permutation testing) # now generate permuted statistics for 100 models permuted.stats<-permute.OSC.PLS(data=scaled.data,y=as.matrix(tmp.y),n=100,ncomp=2,osc.comp=1, progress=FALSE) #look how our model compares to random chance OSC.validate.model(model=mods,perm=permuted.stats) # Q2 represents the in-bag or error for the training data # Xvar the variance in the variables (X) explained or captured in the model # RMSEP is the out-of-bag error # The p-values are from a single-sample t-Test comparing our models # performance parameters (single values) to the permuted distribution (n=100) #next we can estimate the external (for our initial training set) out-of-bag error by conducting model training and testing #generate 100 random splits of the data into 1/3 test and 2/3 trainning sets # and sample equally for each species (strata) train.test.index=test.train.split(nrow(scaled.data),n=100,strata=as.factor(tmp.y)) train.stats<-OSC.PLS.train.test(pls.data = scaled.data,pls.y = tmp.y,train.test.index ,comp=2,OSC.comp=1,cv.scale=TRUE, progress=FALSE) #Compare the ditributions for the models performance statistics permuted values OSC.validate.model(model=mods,perm=permuted.stats,train=train.stats) # This suggests that we have a strong model (far better than random chance) # which capable of correctly predicting the species of the flower. # We can now test this assumption by predicting the species labels # for the test set we excluded from our data before we even started modeling # (an interesting idea would to use the test set to predict the labels for the training set) #reset data scaled.data<-iris[,-5] tmp.group<-iris[,5] # species tmp.y<-matrix(as.numeric(tmp.group),ncol=1) # make numeric matrix #make predictions for the test set mods<-make.OSC.PLS.model(tmp.y,pls.data=scaled.data,comp=2,OSC.comp=1, validation = "LOO", method="oscorespls", cv.scale=TRUE, progress=FALSE,train.test.index=train.test.index.main) #get the true (actual) and predicted values, round them to integers representing the # discreet species labels plot.data=data.frame(predicted = round(mods$predicted.Y[[2]][,1],0),actual= mods$test.y) # note these are numeric but we would prefer to interpret classification of species a class plot.data$predicted<-factor(plot.data$predicted,labels=levels(iris[,5]),levels=1:3) plot.data$actual<-factor(plot.data$actual,labels=levels(iris[,5]),levels=1:3) table(plot.data) # we see that we misclassified two virginica as versicolor, but are otherwise perfect #this is expected based on the similarity in physical properties between these two species pairs(iris[,-5],pch=21,bg=rainbow(nlevels(iris[,5]),alpha=.75)[iris[,5]],upper.panel=NULL) par(xpd=TRUE) legend(.75,1,levels(iris[,5]),fill=rainbow(nlevels(iris[,5]),alpha=.75),bty="n") # we can plot an O-PLS-DA m ================================================ FILE: Demos/Predictive Modeling/O-PLS/OPLS example.Rmd ================================================ --- title: "O-PLS/-DA Modeling" output: html_document: keep_md: yes --- Load data set, and set model x and y. ```{r,message=FALSE, warning=FALSE} #load library now to prevent loading messages Sys.setenv(ALLOW_WGCNA_THREADS=1) suppressPackageStartupMessages(library(WGCNA)) # so annoying #data data(mtcars) #X pls.data<-mtcars[,-1] #y, mpg pls.y<-mtcars[,1,drop=F] #make model opls.results<-make.OSC.PLS.model(pls.y,pls.data, comp=2, OSC.comp=1, validation = "LOO", cv.scale = TRUE, train.test.index=NULL, progress=FALSE) ``` Get 2 latent variables (LVs) and 1 orthogonal LV model stats. ```{r,message=FALSE, warning=FALSE} #extra results as some LV and OSC and print model stats final.opls.results<-get.OSC.model(obj=opls.results,OSC.comp=1) (opls.model.text<-data.frame("Xvar"=c(0,round(cumsum(final.opls.results$Xvar)*100,2)),"Q2"=final.opls.results$Q2,"RMSEP"= final.opls.results$RMSEP) ) ``` Predict mpg values for held out car data and calculate test error (RMSEP). ```{r,message=FALSE, warning=FALSE} #train/test index 2/3 train and 1/3 test train.test.index <- test.train.split(nrow(pls.data), n = 1) #fit model mods<-make.OSC.PLS.model(pls.y,pls.data, comp=2, OSC.comp=1, validation = "LOO", cv.scale = TRUE, train.test.index=train.test.index, progress=FALSE) #view predictions for test data final.opls.results2<-get.OSC.model(obj=mods,OSC.comp=1) fitted<-final.opls.results2$predicted.Y (RMSEP<-(.MSEP(actual=pls.y[train.test.index=="test",],pred=fitted))^.5) ``` Carry out 100 rounds of training and testing cross-validation and get model performance summary. ```{r,message=FALSE, warning=FALSE} #train/test index 100 rounds train.test.index <- test.train.split(nrow(pls.data), n = 100) multi.train.test<-OSC.PLS.train.test(pls.data = pls.data, pls.y = pls.y, train.test.index, comp = mods$total.LVs[1], OSC.comp = max(mods$OSC.LVs), cv.scale = mods$model.description$cv.scale, progress = FALSE) # ... multi.train.test$summary ``` Carry out permutation testing and calculate random chance statistics (null model). ```{r,message=FALSE, warning=FALSE} multi.permute<-permute.OSC.PLS.train.test(pls.data = pls.data, pls.y = pls.y, perm.n = 100, comp = mods$total.LVs[1], OSC.comp=max(mods$OSC.LVs), progress = FALSE, train.test.index = train.test.index) ``` Compare model statistical distrubutions to permuted model performance and calculate proportion of times real model was better then permuted model as a p-value. ```{r,message=FALSE, warning=FALSE} #compare actual to permuted model performance (model.validation<-OSC.validate.model(model = mods, perm = multi.permute, train = multi.train.test,test="perm.test")) ``` Carry out a single round of feature selection select top 4 features and plot results. ```{r,message=FALSE, warning=FALSE} #feature selection opts<-PLS.feature.select(pls.data,pls.scores=final.opls.results$scores[,][,1,drop=F],pls.loadings=final.opls.results$loadings[,][,1,drop=F],pls.weight=final.opls.results$loadings[,][,1,drop=F],plot=FALSE,p.value=0.1,FDR=TRUE,cut.type="number",top=4,separate=FALSE) # make s-plot plus plot.S.plot(obj=opts,return="all") ``` Calculate and compare performance statistics for included and excluded feature models. ```{r,message=FALSE, warning=FALSE} optim<-optimize.OPLS.feature.select(model=opls.results,feature.subset=opts$combined.selection,permute=TRUE,train.test.index,progress=FALSE,test="perm.test") cbind(model=c(rep(c("model","permuted","p-value"),2),"p.value"),optim$summary) ``` Get model stats for decreasing number of model variables using full model loadings calculated above as a gradient. ```{r,message=FALSE, warning=FALSE} #optimize model feature selections filter<-seq(3,ncol(pls.data)-3) # number of variables to keep res<-multi.OPLS.feature.select(model=opls.results,filter=filter,plot=FALSE,OPLSDA=TRUE,train.test.index=train.test.index, test="perm.test", progress=FALSE) # use full model without training split as input plot.multi.OPLS.feature.select(res,objects=c("RMSEP","Q2")) # view results best.OPLS.features(res)[,1:5] # extract best model ``` ================================================ FILE: Demos/Predictive Modeling/O-PLS/OPLS_example.html ================================================ O-PLS/-DA Modeling

Load data set, and set model x and y.

#load library now to prevent loading messages
Sys.setenv(ALLOW_WGCNA_THREADS=1)
suppressPackageStartupMessages(library(WGCNA)) # so annoying
## ==========================================================================
## *
## *  Package WGCNA 1.41.1 loaded.
## *
## *    Important note: It appears that your system supports multi-threading,
## *    but it is not enabled within WGCNA in R. 
## *    To allow multi-threading within WGCNA with all available cores, use 
## *
## *          allowWGCNAThreads()
## *
## *    within R. Use disableWGCNAThreads() to disable threading if necessary.
## *    Alternatively, set the following environment variable on your system:
## *
## *          ALLOW_WGCNA_THREADS=<number_of_processors>
## *
## *    for example 
## *
## *          ALLOW_WGCNA_THREADS=8
## *
## *    To set the environment variable in linux bash shell, type 
## *
## *           export ALLOW_WGCNA_THREADS=8
## *
## *     before running R. Other operating systems or shells will
## *     have a similar command to achieve the same aim.
## *
## ==========================================================================
#data
data(mtcars)

#X
pls.data<-mtcars[,-1]
#y, mpg
pls.y<-mtcars[,1,drop=F]

#make model
opls.results<-make.OSC.PLS.model(pls.y,pls.data,
                        comp=2,
                        OSC.comp=1, 
                        validation = "LOO", 
                        cv.scale = TRUE,
                        train.test.index=NULL,
                        progress=FALSE)

Get 2 latent variables (LVs) and 1 orthogonal LV model stats.

#extra results as some LV and OSC and print model stats
final.opls.results<-get.OSC.model(obj=opls.results,OSC.comp=1)      
(opls.model.text<-data.frame("Xvar"=c(0,round(cumsum(final.opls.results$Xvar)*100,2)),"Q2"=final.opls.results$Q2,"RMSEP"= final.opls.results$RMSEP) )
##         Xvar          Q2    RMSEP
##         0.00 -0.06555671 6.123385
## Comp 1 76.67  0.52660903 4.076258
## Comp 2 88.06  0.74269548 3.004066

Predict mpg values for held out car data and calculate test error (RMSEP).

#train/test index 2/3 train and 1/3 test
train.test.index <- test.train.split(nrow(pls.data), n = 1) 

#fit model 
mods<-make.OSC.PLS.model(pls.y,pls.data,
                        comp=2,
                        OSC.comp=1, 
                        validation = "LOO", 
                        cv.scale = TRUE,
                        train.test.index=train.test.index,
                        progress=FALSE) 

#view predictions for test data
final.opls.results2<-get.OSC.model(obj=mods,OSC.comp=1) 
fitted<-final.opls.results2$predicted.Y
(RMSEP<-(.MSEP(actual=pls.y[train.test.index=="test",],pred=fitted))^.5)
## [1] 3.511172

Carry out 100 rounds of training and testing cross-validation and get model performance summary.

#train/test index 100 rounds
train.test.index <- test.train.split(nrow(pls.data), n = 100) 
multi.train.test<-OSC.PLS.train.test(pls.data = pls.data, pls.y = pls.y, train.test.index, comp = mods$total.LVs[1], OSC.comp = max(mods$OSC.LVs), cv.scale = mods$model.description$cv.scale, progress = FALSE) # ...
multi.train.test$summary
##      Xvar             Q2                 RMSEP           
## [1,] "81.01 +/- 8.35" "0.5406 +/- 0.158" "3.641 +/- 2.34"

Carry out permutation testing and calculate random chance statistics (null model).

multi.permute<-permute.OSC.PLS.train.test(pls.data = pls.data, pls.y = pls.y, perm.n = 100, comp = mods$total.LVs[1], OSC.comp=max(mods$OSC.LVs), progress = FALSE, train.test.index = train.test.index)

Compare model statistical distrubutions to permuted model performance and calculate proportion of times real model was better then permuted model as a p-value.

#compare actual to permuted model performance
(model.validation<-OSC.validate.model(model = mods, perm = multi.permute, train = multi.train.test,test="perm.test"))
##                           Xvar               Q2          RMSEP
## model           81.01 +/- 8.35 0.5406 +/- 0.158 3.641 +/- 2.34
## permuted model 99.93 +/- 0.316 -0.137 +/- 0.131  6.375 +/- 1.3
## p-value                      1         0.009901        0.05941

Carry out a single round of feature selection select top 4 features and plot results.

#feature selection
opts<-PLS.feature.select(pls.data,pls.scores=final.opls.results$scores[,][,1,drop=F],pls.loadings=final.opls.results$loadings[,][,1,drop=F],pls.weight=final.opls.results$loadings[,][,1,drop=F],plot=FALSE,p.value=0.1,FDR=TRUE,cut.type="number",top=4,separate=FALSE)
# make s-plot plus
plot.S.plot(obj=opts,return="all")

## NULL

Calculate and compare performance statistics for included and excluded feature models.

optim<-optimize.OPLS.feature.select(model=opls.results,feature.subset=opts$combined.selection,permute=TRUE,train.test.index,progress=FALSE,test="perm.test") 
cbind(model=c(rep(c("model","permuted","p-value"),2),"p.value"),optim$summary)
##      model      model           Xvar                Q2           RMSEP
## 1    model   selected 98.78 +/- 1.67 0.6627 +/- 0.0698 3.617 +/- 0.764
## 2 permuted   selected      100 +/- 0 -0.1503 +/- 0.075  6.477 +/- 1.07
## 3  p-value   selected              1          0.009901          0.0198
## 4    model   excluded 62.28 +/- 10.7 0.6126 +/- 0.0746  3.453 +/- 0.71
## 5 permuted   excluded 98.42 +/- 4.47 -0.1165 +/- 0.176  6.603 +/- 1.34
## 6  p-value   excluded              1          0.009901          0.0198
## 7  p.value comparison       0.009901            0.2178          0.6931

Get model stats for decreasing number of model variables using full model loadings calculated above as a gradient.

#optimize model feature selections 
filter<-seq(3,ncol(pls.data)-3) # number of variables to keep
res<-multi.OPLS.feature.select(model=opls.results,filter=filter,plot=FALSE,OPLSDA=TRUE,train.test.index=train.test.index, test="perm.test", progress=FALSE) # use full model without training split as input
plot.multi.OPLS.feature.select(res,objects=c("RMSEP","Q2")) # view results

best.OPLS.features(res)[,1:5] # extract best model
##    type vars    model   Xvar        Q2
## 9 model    7 included 92.582 0.4200007
================================================ FILE: Demos/Predictive Modeling/O-PLS/OPLS_example.md ================================================ # O-PLS Modeling Load data set, and set model x and y. ```r #load library now to prevent loading messages Sys.setenv(ALLOW_WGCNA_THREADS=1) suppressPackageStartupMessages(library(WGCNA)) # so annoying ``` ``` ## ========================================================================== ## * ## * Package WGCNA 1.41.1 loaded. ## * ## * Important note: It appears that your system supports multi-threading, ## * but it is not enabled within WGCNA in R. ## * To allow multi-threading within WGCNA with all available cores, use ## * ## * allowWGCNAThreads() ## * ## * within R. Use disableWGCNAThreads() to disable threading if necessary. ## * Alternatively, set the following environment variable on your system: ## * ## * ALLOW_WGCNA_THREADS= ## * ## * for example ## * ## * ALLOW_WGCNA_THREADS=8 ## * ## * To set the environment variable in linux bash shell, type ## * ## * export ALLOW_WGCNA_THREADS=8 ## * ## * before running R. Other operating systems or shells will ## * have a similar command to achieve the same aim. ## * ## ========================================================================== ``` ```r #data data(mtcars) #X pls.data<-mtcars[,-1] #y, mpg pls.y<-mtcars[,1,drop=F] #make model opls.results<-make.OSC.PLS.model(pls.y,pls.data, comp=2, OSC.comp=1, validation = "LOO", cv.scale = TRUE, train.test.index=NULL, progress=FALSE) ``` Get 2 latent variables (LVs) and 1 orthogonal LV model stats. ```r #extra results as some LV and OSC and print model stats final.opls.results<-get.OSC.model(obj=opls.results,OSC.comp=1) (opls.model.text<-data.frame("Xvar"=c(0,round(cumsum(final.opls.results$Xvar)*100,2)),"Q2"=final.opls.results$Q2,"RMSEP"= final.opls.results$RMSEP) ) ``` ``` ## Xvar Q2 RMSEP ## 0.00 -0.06555671 6.123385 ## Comp 1 76.67 0.52660903 4.076258 ## Comp 2 88.06 0.74269548 3.004066 ``` Predict mpg values for held out car data and calculate test error (RMSEP). ```r #train/test index 2/3 train and 1/3 test train.test.index <- test.train.split(nrow(pls.data), n = 1) #fit model mods<-make.OSC.PLS.model(pls.y,pls.data, comp=2, OSC.comp=1, validation = "LOO", cv.scale = TRUE, train.test.index=train.test.index, progress=FALSE) #view predictions for test data final.opls.results2<-get.OSC.model(obj=mods,OSC.comp=1) fitted<-final.opls.results2$predicted.Y (RMSEP<-(.MSEP(actual=pls.y[train.test.index=="test",],pred=fitted))^.5) ``` ``` ## [1] 3.511172 ``` Carry out 100 rounds of training and testing cross-validation and get model performance summary. ```r #train/test index 100 rounds train.test.index <- test.train.split(nrow(pls.data), n = 100) multi.train.test<-OSC.PLS.train.test(pls.data = pls.data, pls.y = pls.y, train.test.index, comp = mods$total.LVs[1], OSC.comp = max(mods$OSC.LVs), cv.scale = mods$model.description$cv.scale, progress = FALSE) # ... multi.train.test$summary ``` ``` ## Xvar Q2 RMSEP ## [1,] "81.01 +/- 8.35" "0.5406 +/- 0.158" "3.641 +/- 2.34" ``` Carry out permutation testing and calculate random chance statistics (null model). ```r multi.permute<-permute.OSC.PLS.train.test(pls.data = pls.data, pls.y = pls.y, perm.n = 100, comp = mods$total.LVs[1], OSC.comp=max(mods$OSC.LVs), progress = FALSE, train.test.index = train.test.index) ``` Compare model statistical distrubutions to permuted model performance and calculate proportion of times real model was better then permuted model as a p-value. ```r #compare actual to permuted model performance (model.validation<-OSC.validate.model(model = mods, perm = multi.permute, train = multi.train.test,test="perm.test")) ``` ``` ## Xvar Q2 RMSEP ## model 81.01 +/- 8.35 0.5406 +/- 0.158 3.641 +/- 2.34 ## permuted model 99.93 +/- 0.316 -0.137 +/- 0.131 6.375 +/- 1.3 ## p-value 1 0.009901 0.05941 ``` Carry out a single round of feature selection select top 4 features and plot results. ```r #feature selection opts<-PLS.feature.select(pls.data,pls.scores=final.opls.results$scores[,][,1,drop=F],pls.loadings=final.opls.results$loadings[,][,1,drop=F],pls.weight=final.opls.results$loadings[,][,1,drop=F],plot=FALSE,p.value=0.1,FDR=TRUE,cut.type="number",top=4,separate=FALSE) # make s-plot plus plot.S.plot(obj=opts,return="all") ``` ![](OPLS_example_files/figure-html/unnamed-chunk-7-1.png) ``` ## NULL ``` Calculate and compare performance statistics for included and excluded feature models. ```r optim<-optimize.OPLS.feature.select(model=opls.results,feature.subset=opts$combined.selection,permute=TRUE,train.test.index,progress=FALSE,test="perm.test") cbind(model=c(rep(c("model","permuted","p-value"),2),"p.value"),optim$summary) ``` ``` ## model model Xvar Q2 RMSEP ## 1 model selected 98.78 +/- 1.67 0.6627 +/- 0.0698 3.617 +/- 0.764 ## 2 permuted selected 100 +/- 0 -0.1503 +/- 0.075 6.477 +/- 1.07 ## 3 p-value selected 1 0.009901 0.0198 ## 4 model excluded 62.28 +/- 10.7 0.6126 +/- 0.0746 3.453 +/- 0.71 ## 5 permuted excluded 98.42 +/- 4.47 -0.1165 +/- 0.176 6.603 +/- 1.34 ## 6 p-value excluded 1 0.009901 0.0198 ## 7 p.value comparison 0.009901 0.2178 0.6931 ``` Get model stats for decreasing number of model variables using full model loadings calculated above as a gradient. ```r #optimize model feature selections filter<-seq(3,ncol(pls.data)-3) # number of variables to keep res<-multi.OPLS.feature.select(model=opls.results,filter=filter,plot=FALSE,OPLSDA=TRUE,train.test.index=train.test.index, test="perm.test", progress=FALSE) # use full model without training split as input plot.multi.OPLS.feature.select(res,objects=c("RMSEP","Q2")) # view results ``` ![](OPLS_example_files/figure-html/unnamed-chunk-9-1.png) ```r best.OPLS.features(res)[,1:5] # extract best model ``` ``` ## type vars model Xvar Q2 ## 9 model 7 included 92.582 0.4200007 ``` ================================================ FILE: Demos/Predictive Modeling/Single and multiple Y O-PLS modeling/OPLS example.Rmd ================================================ Orthogonal Partial Least Squares (OPLS) ======================================================== #### This is an example of OPLS modeling in R. See here for [more OPLS options](https://raw.github.com/dgrapov/devium/master/R/Devium%20PLS%20%20and%20OPLS.r) or try [PLS](http://cran.r-project.org/web/packages/pls/index.html). ```{r,include=FALSE,} source("http://pastebin.com/raw.php?i=UyDBTA57") # source Devium ``` #### Generate some random data and Y's. ```{r} set.seed(1234) data<-matrix(rnorm(10000,0,1),nrow=100, ncol=100) simple.y<-matrix(rep(1:2,50),,1) complex.y<-matrix(sample(1:2,400,replace=T),,4) ``` #### Make exploratory model. Fit 10 latent variable (LVs) and 10 orthogonal latent variables (OLVs). ```{r,hide=TRUE,message=FALSE} #scale data scaled.data<-data.frame(scale(data,scale=T,center=T)) comp<-ocomp<-5 # maximum number of latent variables (LVs) pls.y<-simple.y mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=comp,OSC.comp=ocomp,validation = "LOO",method="oscorespls",cv.scale=T) ``` #### View root mean squared error of prediction for various number of OLV models. Note X-axis reffers to total number of OLVs not LVs, hence the start at 0. ```{r,message=FALSE} plot.OSC.results(obj=mods1,plot="RMSEP",groups=group) ``` #### View scores for various models. ```{r,message=FALSE} #create factor to visualize groups group<-factor(join.columns(pls.y))#visualize levels of y plot.OSC.results(obj=mods1,plot="scores",groups=group) ``` #### View variable loadings for various models. ```{r,message=FALSE} #create factor to visualize groups plot.OSC.results(obj=mods1,plot="loadings") ``` #### Get optimal LV/OLV suggestions. This becaomes very handy with multiple Ys. ```{r,hide=TRUE} #fit 1:limit LV/OLV models to overview optimal LV and OLV optimal.model<-optimize.OPLS(max.LV=comp, # max LV tolerance =0.01, #tolerance for accepting higher error models but which are simpler pls.y=pls.y,pls.data=scaled.data, # y and data validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options progress=FALSE) # see pls for theses options ``` #### View optimization suggestions. ```{r} optimal.model ``` ##### ```tolerance``` is used to accept higher RMSEP but simpler models. #### Build optimized model based on optimal.model suggestions. ```{r,hide=TRUE} mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T) ``` #### Get all model information. ```{r} final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) ``` #### View model scores. ```{r} group<-factor(join.columns(pls.y))#visualize levels of y plot.PLS.results(obj=final,plot="scores",groups=group) ``` #### The next step for modeling would be to validate, but I will instead show complex. y modeling. #### Complex multiple column Y. ```{r,hide=TRUE} # make exploratory model to determine orthogonal LV (OLV) number comp<-6 # maximum number of latent variables (LVs) pls.y<-complex.y #fit 1:limit LV/OLV models to overview optimal LV and OLV optimal.model<-optimize.OPLS(max.LV=comp, # max LV tolerance =0.01, #tolerance for accepting higher error models but which are simpler pls.y=pls.y,pls.data=scaled.data, # y and data validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options progress=FALSE) # see pls for theses options ``` #### View suggestions. ```{r} optimal.model ``` #### Build optimized model and plot scores. ```{r,hide=TRUE} mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T) final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) # get all model information #view model scores group<-factor(join.columns(pls.y))#visualize levels of y plot.PLS.results(obj=final,plot="scores",groups=group) ``` #### An alternative to modeling a multiple Ys is to define a single Y based on the multiple columns. This will try to organize all group scores in one dimension (LV1). ```{r,hide=TRUE} pls.y<-matrix(as.numeric(as.factor(join.columns(complex.y))),,1) # create numeric representation #fit 1:limit LV/OLV models to overview optimal LV and OLV optimal.model<-optimize.OPLS(max.LV=comp, # max LV tolerance =0.01, #tolerance for accepting higher error models but which are simpler pls.y=pls.y,pls.data=scaled.data, # y and data validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options progress=FALSE) ``` #### View suggestions. ```{r} optimal.model # currently single LV models will cause an error so limit LV minimium to 2 if(optimal.model$LV==1){optimal.model$LV<-2} ``` #### Build optimized model based on optimal.model suggestions. ```{r,hide=TRUE} mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T) final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) # get all model information ``` #### View model scores. ```{r} group<-factor(join.columns(pls.y))#visualize levels of y plot.PLS.results(obj=final,plot="scores",groups=group) ``` ================================================ FILE: Demos/Predictive Modeling/Single and multiple Y O-PLS modeling/OPLS example.html ================================================ Orthogonal Partial Least Squares (OPLS)

Orthogonal Partial Least Squares (OPLS)

This is an example of OPLS modeling in R. See here for more OPLS options or try PLS.

Generate some random data and Y's.

set.seed(1234)
data<-matrix(rnorm(10000,0,1),nrow=100, ncol=100)
simple.y<-matrix(rep(1:2,50),,1)
complex.y<-matrix(sample(1:2,400,replace=T),,4)

Make exploratory model. Fit 10 latent variable (LVs) and 10 orthogonal latent variables (OLVs).

#scale data 
scaled.data<-data.frame(scale(data,scale=T,center=T)) 

comp<-ocomp<-5 # maximum number of latent variables (LVs)
pls.y<-simple.y

mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=comp,OSC.comp=ocomp,validation = "LOO",method="oscorespls",cv.scale=T)

View root mean squared error of prediction for various number of OLV models. Note X-axis reffers to total number of OLVs not LVs, hence the start at 0.

plot.OSC.results(obj=mods1,plot="RMSEP",groups=group)

plot of chunk unnamed-chunk-4

View scores for various models.

#create factor to visualize groups
group<-factor(join.columns(pls.y))#visualize levels of y 
plot.OSC.results(obj=mods1,plot="scores",groups=group)

plot of chunk unnamed-chunk-5

View variable loadings for various models.

#create factor to visualize groups
plot.OSC.results(obj=mods1,plot="loadings")

plot of chunk unnamed-chunk-6

Get optimal LV/OLV suggestions. This becaomes very handy with multiple Ys.

#fit 1:limit LV/OLV models to overview optimal LV and OLV
optimal.model<-optimize.OPLS(max.LV=comp, # max LV
                        tolerance =0.01, #tolerance for accepting higher error models but which are simpler
                            pls.y=pls.y,pls.data=scaled.data, # y and data
                            validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options
                            progress=FALSE) # see pls for theses options

View optimization suggestions.

optimal.model

## $best
##   RMSEP.1 RMSEP.2 LV OLV pls.y delta.tmp.min.
## 9  0.4724  0.4704  2   2     1              0
## 
## $LV
## [1] 2
## 
## $OLV
## [1] 2
tolerance is used to accept higher RMSEP but simpler models.

Build optimized model based on optimal.model suggestions.

mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T)

Get all model information.

final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) 

## Error: 'names' attribute [1] must be the same length as the vector [0]

View model scores.

group<-factor(join.columns(pls.y))#visualize levels of y 
plot.PLS.results(obj=final,plot="scores",groups=group)

## Error: object 'final' not found

The next step for modeling would be to validate, but I will instead show complex. y modeling.

Complex multiple column Y.

# make exploratory model to determine orthogonal LV (OLV) number 
comp<-6 # maximum number of latent variables (LVs)
pls.y<-complex.y
#fit 1:limit LV/OLV models to overview optimal LV and OLV
optimal.model<-optimize.OPLS(max.LV=comp, # max LV
                        tolerance =0.01, #tolerance for accepting higher error models but which are simpler
                            pls.y=pls.y,pls.data=scaled.data, # y and data
                            validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options
                            progress=FALSE)  # see pls for theses options

View suggestions.

optimal.model

## $best
##     RMSEP.1 RMSEP.2 LV OLV pls.y delta.tmp.min.
## 54   0.4541  0.4528  5   2     1      5.140e-03
## 78   0.4531  0.4521  5   3     1      4.435e-03
## 102  0.4487  0.4476  5   4     1      0.000e+00
## 126  0.4488  0.4478  5   5     1      1.326e-04
## 30   0.4541  0.4528  2   2     2      3.125e-03
## 50   0.4506  0.4497  4   2     2      6.253e-05
## 70   0.4520  0.4511  4   3     2      1.458e-03
## 90   0.4506  0.4496  4   4     2      0.000e+00
## 14   0.4703  0.4693  1   1     3      2.804e-03
## 75   0.4692  0.4676  4   3     3      1.105e-03
## 95   0.4680  0.4665  4   4     3      0.000e+00
## 66   0.4778  0.4763  5   2     3      9.788e-03
## 114  0.4726  0.4711  5   4     3      4.601e-03
## 138  0.4700  0.4686  5   5     3      2.048e-03
## 16   0.4911  0.4910  1   1     4      0.000e+00
## 24   0.4989  0.4981  2   1     4      7.131e-03
## 36   0.4995  0.4987  2   2     4      7.718e-03
## 
## $LV
## [1] 5
## 
## $OLV
## [1] 4

Build optimized model and plot scores.

mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T)
final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) # get all model information

## Error: 'names' attribute [1] must be the same length as the vector [0]

#view model scores
group<-factor(join.columns(pls.y))#visualize levels of y 
plot.PLS.results(obj=final,plot="scores",groups=group)

## Error: object 'final' not found

An alternative to modeling a multiple Ys is to define a single Y based on the multiple columns. This will try to organize all group scores in one dimension (LV1).

pls.y<-matrix(as.numeric(as.factor(join.columns(complex.y))),,1) # create numeric representation
#fit 1:limit LV/OLV models to overview optimal LV and OLV
optimal.model<-optimize.OPLS(max.LV=comp, # max LV
                        tolerance =0.01, #tolerance for accepting higher error models but which are simpler
                            pls.y=pls.y,pls.data=scaled.data, # y and data
                            validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options
                            progress=FALSE)

View suggestions.

optimal.model

## $best
##   RMSEP.1 RMSEP.2 LV OLV pls.y delta.tmp.min.
## 4   4.103    4.09  1   1     1              0
## 
## $LV
## [1] 1
## 
## $OLV
## [1] 1

# currently single LV models will cause an error so limit LV minimium to 2
if(optimal.model$LV==1){optimal.model$LV<-2}

Build optimized model based on optimal.model suggestions.

mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T)
final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) # get all model information

## Error: 'names' attribute [1] must be the same length as the vector [0]

View model scores.

group<-factor(join.columns(pls.y))#visualize levels of y 
plot.PLS.results(obj=final,plot="scores",groups=group) 

## Error: object 'final' not found
================================================ FILE: Demos/Predictive Modeling/Single and multiple Y O-PLS modeling/OPLS example.md ================================================ Orthogonal Partial Least Squares (OPLS) ======================================================== #### This is an example of OPLS modeling in R. See here for [more OPLS options](https://raw.github.com/dgrapov/devium/master/R/Devium%20PLS%20%20and%20OPLS.r) or try [PLS](http://cran.r-project.org/web/packages/pls/index.html). #### Generate some random data and Y's. ```r set.seed(1234) data<-matrix(rnorm(10000,0,1),nrow=100, ncol=100) simple.y<-matrix(rep(1:2,50),,1) complex.y<-matrix(sample(1:2,400,replace=T),,4) ``` #### Make exploratory model. Fit 10 latent variable (LVs) and 10 orthogonal latent variables (OLVs). ```r #scale data scaled.data<-data.frame(scale(data,scale=T,center=T)) comp<-ocomp<-5 # maximum number of latent variables (LVs) pls.y<-simple.y mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=comp,OSC.comp=ocomp,validation = "LOO",method="oscorespls",cv.scale=T) ``` #### View root mean squared error of prediction for various number of OLV models. Note X-axis reffers to total number of OLVs not LVs, hence the start at 0. ```r plot.OSC.results(obj=mods1,plot="RMSEP",groups=group) ``` ![plot of chunk unnamed-chunk-4](figure/unnamed-chunk-4.png) #### View scores for various models. ```r #create factor to visualize groups group<-factor(join.columns(pls.y))#visualize levels of y plot.OSC.results(obj=mods1,plot="scores",groups=group) ``` ![plot of chunk unnamed-chunk-5](figure/unnamed-chunk-5.png) #### View variable loadings for various models. ```r #create factor to visualize groups plot.OSC.results(obj=mods1,plot="loadings") ``` ![plot of chunk unnamed-chunk-6](figure/unnamed-chunk-6.png) #### Get optimal LV/OLV suggestions. This becaomes very handy with multiple Ys. ```r #fit 1:limit LV/OLV models to overview optimal LV and OLV optimal.model<-optimize.OPLS(max.LV=comp, # max LV tolerance =0.01, #tolerance for accepting higher error models but which are simpler pls.y=pls.y,pls.data=scaled.data, # y and data validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options progress=FALSE) # see pls for theses options ``` #### View optimization suggestions. ```r optimal.model ``` ``` ## $best ## RMSEP.1 RMSEP.2 LV OLV pls.y delta.tmp.min. ## 9 0.4724 0.4704 2 2 1 0 ## ## $LV ## [1] 2 ## ## $OLV ## [1] 2 ``` ##### ```tolerance``` is used to accept higher RMSEP but simpler models. #### Build optimized model based on optimal.model suggestions. ```r mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T) ``` #### Get all model information. ```r final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) ``` ``` ## Error: 'names' attribute [1] must be the same length as the vector [0] ``` #### View model scores. ```r group<-factor(join.columns(pls.y))#visualize levels of y plot.PLS.results(obj=final,plot="scores",groups=group) ``` ``` ## Error: object 'final' not found ``` #### The next step for modeling would be to validate, but I will instead show complex. y modeling. #### Complex multiple column Y. ```r # make exploratory model to determine orthogonal LV (OLV) number comp<-6 # maximum number of latent variables (LVs) pls.y<-complex.y #fit 1:limit LV/OLV models to overview optimal LV and OLV optimal.model<-optimize.OPLS(max.LV=comp, # max LV tolerance =0.01, #tolerance for accepting higher error models but which are simpler pls.y=pls.y,pls.data=scaled.data, # y and data validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options progress=FALSE) # see pls for theses options ``` #### View suggestions. ```r optimal.model ``` ``` ## $best ## RMSEP.1 RMSEP.2 LV OLV pls.y delta.tmp.min. ## 54 0.4541 0.4528 5 2 1 5.140e-03 ## 78 0.4531 0.4521 5 3 1 4.435e-03 ## 102 0.4487 0.4476 5 4 1 0.000e+00 ## 126 0.4488 0.4478 5 5 1 1.326e-04 ## 30 0.4541 0.4528 2 2 2 3.125e-03 ## 50 0.4506 0.4497 4 2 2 6.253e-05 ## 70 0.4520 0.4511 4 3 2 1.458e-03 ## 90 0.4506 0.4496 4 4 2 0.000e+00 ## 14 0.4703 0.4693 1 1 3 2.804e-03 ## 75 0.4692 0.4676 4 3 3 1.105e-03 ## 95 0.4680 0.4665 4 4 3 0.000e+00 ## 66 0.4778 0.4763 5 2 3 9.788e-03 ## 114 0.4726 0.4711 5 4 3 4.601e-03 ## 138 0.4700 0.4686 5 5 3 2.048e-03 ## 16 0.4911 0.4910 1 1 4 0.000e+00 ## 24 0.4989 0.4981 2 1 4 7.131e-03 ## 36 0.4995 0.4987 2 2 4 7.718e-03 ## ## $LV ## [1] 5 ## ## $OLV ## [1] 4 ``` #### Build optimized model and plot scores. ```r mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T) final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) # get all model information ``` ``` ## Error: 'names' attribute [1] must be the same length as the vector [0] ``` ```r #view model scores group<-factor(join.columns(pls.y))#visualize levels of y plot.PLS.results(obj=final,plot="scores",groups=group) ``` ``` ## Error: object 'final' not found ``` #### An alternative to modeling a multiple Ys is to define a single Y based on the multiple columns. This will try to organize all group scores in one dimension (LV1). ```r pls.y<-matrix(as.numeric(as.factor(join.columns(complex.y))),,1) # create numeric representation #fit 1:limit LV/OLV models to overview optimal LV and OLV optimal.model<-optimize.OPLS(max.LV=comp, # max LV tolerance =0.01, #tolerance for accepting higher error models but which are simpler pls.y=pls.y,pls.data=scaled.data, # y and data validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options progress=FALSE) ``` #### View suggestions. ```r optimal.model ``` ``` ## $best ## RMSEP.1 RMSEP.2 LV OLV pls.y delta.tmp.min. ## 4 4.103 4.09 1 1 1 0 ## ## $LV ## [1] 1 ## ## $OLV ## [1] 1 ``` ```r # currently single LV models will cause an error so limit LV minimium to 2 if(optimal.model$LV==1){optimal.model$LV<-2} ``` #### Build optimized model based on optimal.model suggestions. ```r mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T) final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) # get all model information ``` ``` ## Error: 'names' attribute [1] must be the same length as the vector [0] ``` #### View model scores. ```r group<-factor(join.columns(pls.y))#visualize levels of y plot.PLS.results(obj=final,plot="scores",groups=group) ``` ``` ## Error: object 'final' not found ``` ================================================ FILE: Demos/Predictive Modeling/Single and multiple Y O-PLS modeling/OPLS example.r ================================================ # OPLS example source("http://pastebin.com/raw.php?i=JVyTrYRD") # source Devium #generate some random data and Y's set.seed(1234) data<-matrix(rnorm(10000,0,1),nrow=100, ncol=100) simple.y<-matrix(rep(1:2,50),,1) complex.y<-matrix(sample(1:2,400,replace=T),,4) #scale data scaled.data<-data.frame(scale(data,scale=T,center=T)) # make exploratory model to determine orthogonal LV (OLV) number comp<-3 # maximum number of latent variables (LVs) pls.y<-simple.y #### Make exploratory model. Fit 10 latent variable (LVs) and 10 orthogonal latent variables (OLVs). #scale data scaled.data<-data.frame(scale(data,scale=T,center=T)) comp<-ocomp<-10 # maximum number of latent variables (LVs) pls.y<-simple.y mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=comp,OSC.comp=ocomp,validation = "LOO",method="oscorespls",cv.scale=F) #### view root mean squared error of prediction for various number of OLV models. Note X-axis reffers to total number of OLVs not LVs, hence the start at 0. plot.OSC.results(obj=mods1,plot="RMSEP",groups=group) #fit 1:limit LV/OLV models to overview optimal LV and OLV optimal.model<-optimize.OPLS(max.LV=comp, # max LV tolerance =0.01, #tolerance for accepting higher error models but which are simpler pls.y=pls.y,pls.data=scaled.data, # y and data validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options progress=FALSE) #view suggestions optimal.model # suggests 3 components all OLV... #build optimized model based on optimal.model suggestions mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T) final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) # get all model information # view RMSEP plot.PLS.results(obj=final,plot="RMSEP",groups=group) #view model scores group<-factor(join.columns(pls.y))#visualize levels of y plot.PLS.results(obj=final,plot="scores",groups=group) #the next step for modeling would be to validate, but I will instead show complex. y modeling # make exploratory model to determine orthogonal LV (OLV) number comp<-6 # maximum number of latent variables (LVs) pls.y<-complex.y #fit 1:limit LV/OLV models to overview optimal LV and OLV optimal.model<-optimize.OPLS(max.LV=comp, # max LV tolerance =0.01, #tolerance for accepting higher error models but which are simpler pls.y=pls.y,pls.data=scaled.data, # y and data validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options progress=FALSE) #view suggestions optimal.model # suggests 3 components all OLV... #build optimized model based on optimal.model suggestions mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T) final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) # get all model information #view model scores group<-factor(join.columns(pls.y))#visualize levels of y plot.PLS.results(obj=final,plot="scores",groups=group) #an alternative to modeling a multiple Ys is to define a single Y based on the multiple columns # this will try to organize all group scores in one dimension (LV1) pls.y<-matrix(as.numeric(as.factor(join.columns(complex.y))),,1) # create numeric representation #fit 1:limit LV/OLV models to overview optimal LV and OLV optimal.model<-optimize.OPLS(max.LV=comp, # max LV tolerance =0.01, #tolerance for accepting higher error models but which are simpler pls.y=pls.y,pls.data=scaled.data, # y and data validation = "LOO",method="oscorespls",cv.scale=F,# see pls for theses options progress=FALSE) #view suggestions optimal.model # currently single LV models will caause an error so limit LV minimium to 2 if(optimal.model$LV==1){optimal.model$LV<-2} #build optimized model based on optimal.model suggestions mods1<-OSC.correction(progress=FALSE,pls.y=pls.y,pls.data=scaled.data,comp=optimal.model$LV,OSC.comp=optimal.model$OLV,validation = "LOO",method="oscorespls",cv.scale=T) final<-get.OSC.model(obj=mods1,OSC.comp=optimal.model$OLV) # get all model information #view model scores group<-factor(join.columns(pls.y))#visualize levels of y plot.PLS.results(obj=final,plot="scores",groups=group) ================================================ FILE: Demos/R Programming/R tutorial 1.R ================================================ #------------------------------------ # getting started with R lesson 1 # by Dmitry Grapov #------------------------------------ #------------------------------------ # GOALS: # 1) limited overview of basic objects # 2) plot some data # 3) create a data summary # 4) focus on bare bones (try to avid convenience fxns for now) to learn basic concepts #------------------------------------ #R tutorial # this is a comment R doesn't interpret this 10 # works a # what happens here "a" # works a<-10 # assignment b<-a+1 # using a pre assigned variable #Arithmetic a + b #addition a - b #subtraction a * b #multiplication a / b #division a^b #exponentiation a %/% b #integer division a %% b #modulo (remainder) b%%a makes more sense #Relational a==b #is a equal to b a>b # a is greater than b, use < for less than a>=b # a is greater than or equal to #Logical ! #not a!=b # could also do !a==b & #and | #or && #sequential and || #sequential or #lets create a new variable to test the logical operators x<-1:10 # note ':' creates a sequence x!=a # notice how the shorter a is recycled (id<-!x>=a) # notice we assigned the results to id and use () to print this x>a|xa&x=2 x[c(1,2,6:8)] # can use c() to get many specific elements x[id] # we can also use a logical to get our object # lets find out more about x str(x) #structure, x is an integer vector (one dimension) class(id) # is logical length(x) # length # next lets load some real data to experiment with data(iris) # we use a function named data to load the iris data str(iris) # structure #notice we have 2 dimensions now, rows and columns # now subset the object as object[rows,columns] iris[1:5,3:4] species<-iris$Species # in data.frames and list we can also reference columns with '$' species<-iris[,"Species"] # or by name #lets make some plots plot(iris[,1:2]) # plot the first 2 columns help(plot) #see what other arguments plot could take plot(iris[,1],species) plot(iris[,1]~species) # plot is different because species is factor and formula notation "~" is used plot(iris[,1]~species, col =c("red","green","blue")) # add a color plot(iris[,1]~species, col =c("red","green","blue"),ylab=colnames(iris)[1]) # and label, use function colnames() to get column names and take the first columns name plot(iris) # because we gave the whole data frame R calls pairs() and creates a scatterplot matrix color<-c("red","green","blue")[species] # we can use the factor to subset our colors to create a color for each point plot(iris, pch=21,bg=color, main="My Awesome Plot!") # here we give custom point shape 'pch', border 'col' and inner color 'bg' as well as a title 'main' #next lets experiment with getting summary statics mean(iris) # we want to get the mean but giving the whole data.frame with the factor does not make sense? mean(iris[,1]) # this works mean(iris[,"Species"]) # this was the issue, it is not numeric (we could coerce 'as.numeric' but why?) mean(iris[iris$Species=="setosa",1]) # here we subset the rows to only get values for the species setosa and return the mean for column 1 for this group #get all column means for setosa apply(iris[iris$Species=="setosa",1:4],2,mean) #here we use function apply to for each column use the function mean (see help(apply)) # we also removed species because we know mean wont work on this and also causes bad behaviour for other columns # now lets get the means for each species big.l<-split(iris[,1:4], iris$Species) # create a list holding a data.frame for each level of the species (res<-lapply(big.l,apply,2,mean)) # get means for each species and variable data.frame(res) # get results combined t(data.frame(res)) # transpose results ================================================ FILE: Demos/R Programming/R tutorial 1.html ================================================ Getting Started with R Lesson 1

Getting Started with R Lesson 1

by Dmitry Grapov

GOALS:

* getting started

* basic operators

* plotting

* creating data summaries

For the following examples we will focus on the bare bones to help us learn basic concepts

Getting started

# this is a comment R doesn't interpret this
10  # works

## [1] 10

# a # remove comment and try to type a

R knows what 10 is, it is a number which has been defined in the R environment. However a has not been defined and R throws an error when trying to print it to the console.

"a"  # works

## [1] "a"

a <- 10  # assignment
b <- a + 1  # using a pre assigned variable

When we give a we define that this object is as a character vector “a” which R understands how to print to screen. We can assign 10 to something R knows about like for instance 10. Having done this we can now use 10 as a variable to do whatever we want.

Arithmetic operators

a + b  #addition

## [1] 21

a - b  #subtraction

## [1] -1

a * b  #multiplication

## [1] 110

a/b  #division

## [1] 0.9091

a^b  #exponentiation

## [1] 1e+11

a%/%b  #integer division

## [1] 0

a%%b  #modulo (remainder) b%%a makes more sense

## [1] 10

Relational operators

a == b  #is a equal to b

## [1] FALSE

a > b  # a is greater than b, use < for less than

## [1] FALSE

a >= b  # a is greater than or equal to

## [1] FALSE

Logical operators

! #not
a!=b # is a equal to be? could also do !a==b    

## [1] FALSE

#& #and
# | #or
# && #sequential and
# || #sequential or

Lets create a new variable to test the logical operators

x <- 1:10  # note ':' creates a sequence
x != a  # notice how the shorter a is recycled

##  [1]  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE FALSE

(id <- !x >= a)  # notice we assigned the results to id and use () to print this

##  [1]  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE FALSE

x > a | x < b  # check if any criteria specified is TRUE

##  [1] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE

x > a & x < b  # check if all criteria specified is TRUE

##  [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE

Indexing

a[1]  # get the first 'unit' of a

## [1] 10

a[2]  #  trying to reference something which doesn't exist, NA stands for missing value

## [1] NA

x[2]  # this works because x has length >=2

## [1] 2

x[c(1, 2, 6:8)]  # can use c() to get many specific elements

## [1] 1 2 6 7 8

x[id]  # we can also use a logical to get our object

## [1] 1 2 3 4 5 6 7 8 9

Object properties

str(x)  #structure, x is an integer vector (one dimension)

##  int [1:10] 1 2 3 4 5 6 7 8 9 10

class(id)  # is logical

## [1] "logical"

length(x)  # length

## [1] 10

Next lets load some real data to experiment with

data(iris)  # we use a function named data to load the iris data
str(iris)  # structure

## 'data.frame':    150 obs. of  5 variables:
##  $ Sepal.Length: num  5.1 4.9 4.7 4.6 5 5.4 4.6 5 4.4 4.9 ...
##  $ Sepal.Width : num  3.5 3 3.2 3.1 3.6 3.9 3.4 3.4 2.9 3.1 ...
##  $ Petal.Length: num  1.4 1.4 1.3 1.5 1.4 1.7 1.4 1.5 1.4 1.5 ...
##  $ Petal.Width : num  0.2 0.2 0.2 0.2 0.2 0.4 0.3 0.2 0.2 0.1 ...
##  $ Species     : Factor w/ 3 levels "setosa","versicolor",..: 1 1 1 1 1 1 1 1 1 1 ...

# notice we have 2 dimensions now, rows and columns now subset the object as
# object[rows,columns]
iris[1:5, 3:4]

##   Petal.Length Petal.Width
## 1          1.4         0.2
## 2          1.4         0.2
## 3          1.3         0.2
## 4          1.5         0.2
## 5          1.4         0.2

species <- iris$Species  # in data.frames and list we can also reference columns with '$'
species <- iris[, "Species"]  # or by name

Lets make some plots

plot(iris[, 1:2])  # plot the first 2 columns

plot of chunk unnamed-chunk-10 

help(plot)  #see what other arguments plot could take
plot(iris[, 1], species)

plot of chunk unnamed-chunk-10 

plot(iris[, 1] ~ species)  # plot is different because species is factor and formula notation '~' is used

plot of chunk unnamed-chunk-10 

plot(iris[, 1] ~ species, col = c("red", "green", "blue"))  # add a color

plot of chunk unnamed-chunk-10 

plot(iris[, 1] ~ species, col = c("red", "green", "blue"), ylab = colnames(iris)[1])  # and label, use function colnames() to get column names and take the first columns name

plot of chunk unnamed-chunk-10 

plot(iris)  # because we gave the whole data frame R calls pairs() and creates a scatterplot matrix

plot of chunk unnamed-chunk-10 

color <- c("red", "green", "blue")[species]  # we can use the factor to subset our colors to create a color for each point
plot(iris, pch = 21, bg = color, main = "My Awesome Plot!")  # here we give custom point shape 'pch', border 'col' and inner color 'bg' as well as a title 'main'

plot of chunk unnamed-chunk-10

Next lets experiment with getting summary statics

mean(iris)  # we want to get the mean but giving the whole data.frame with the factor does not make sense?

## [1] NA

mean(iris[, 1])  # this works

## [1] 5.843

mean(iris[, "Species"])  # this was the issue, it is not numeric (we could coerce 'as.numeric' but why?)

## [1] NA

mean(iris[iris$Species == "setosa", 1])  # here we subset the rows to only get values for the species setosa and return the mean for column 1 for this group

## [1] 5.006

Next lets get all column means for the species setosa

apply(iris[iris$Species == "setosa", 1:4], 2, mean)  #here we use function apply to for each column  use the function mean (see help(apply)) 

## Sepal.Length  Sepal.Width Petal.Length  Petal.Width 
##        5.006        3.428        1.462        0.246

# we also removed species because we know mean wont work on this and also
# causes bad behaviour for other columns

Now lets get the means for each species

big.l <- split(iris[, 1:4], iris$Species)  # create a list holding a data.frame for each level of the species
(res <- lapply(big.l, apply, 2, mean))  # get means for each species and variable

## $setosa
## Sepal.Length  Sepal.Width Petal.Length  Petal.Width 
##        5.006        3.428        1.462        0.246 
## 
## $versicolor
## Sepal.Length  Sepal.Width Petal.Length  Petal.Width 
##        5.936        2.770        4.260        1.326 
## 
## $virginica
## Sepal.Length  Sepal.Width Petal.Length  Petal.Width 
##        6.588        2.974        5.552        2.026

data.frame(res)  # get results combined

##              setosa versicolor virginica
## Sepal.Length  5.006      5.936     6.588
## Sepal.Width   3.428      2.770     2.974
## Petal.Length  1.462      4.260     5.552
## Petal.Width   0.246      1.326     2.026

t(data.frame(res))  # transpose results 

##            Sepal.Length Sepal.Width Petal.Length Petal.Width
## setosa            5.006       3.428        1.462       0.246
## versicolor        5.936       2.770        4.260       1.326
## virginica         6.588       2.974        5.552       2.026

Using this basic syntax we can exucute any custom function that we want (more to come later on R libraries to this this in an elegant and efficient way).

Lets define a custom function to return the formatted mean and standard deviation, and then use it in the code above.

mean.sd <- function(x) {
    paste0(round(c(mean(x), sd(x)), 1), collapse = " +/- ")  # our custom function

}
t(sapply(big.l, apply, 2, mean.sd))

##            Sepal.Length  Sepal.Width   Petal.Length  Petal.Width  
## setosa     "5 +/- 0.4"   "3.4 +/- 0.4" "1.5 +/- 0.2" "0.2 +/- 0.1"
## versicolor "5.9 +/- 0.5" "2.8 +/- 0.3" "4.3 +/- 0.5" "1.3 +/- 0.2"
## virginica  "6.6 +/- 0.6" "3 +/- 0.3"   "5.6 +/- 0.6" "2 +/- 0.3"
================================================ FILE: Demos/R Programming/R tutorial 1.md ================================================ # Getting Started with R Lesson 1 ### by Dmitry Grapov ### GOALS: ### * [getting started](#basics) ### * [basic operators](#operators) ### * [plotting](#plotting) ### * [creating data summaries](#summary) ### For the following examples we will focus on the bare bones to help us learn basic concepts ### Getting started ```r # this is a comment R doesn't interpret this 10 # works ``` ``` ## [1] 10 ``` ```r # a # remove comment and try to type a ``` R knows what ``10`` is, it is a number which has been defined in the R environment. However a has not been defined and R throws an error when trying to print it to the console. ```r "a" # works ``` ``` ## [1] "a" ``` ```r a <- 10 # assignment b <- a + 1 # using a pre assigned variable ``` When we give ``a`` we define that this object is as a character vector "a" which R understands how to print to screen. We can assign ``10`` to something R knows about like for instance ``10``. Having done this we can now use ``10`` as a variable to do whatever we want. ### Arithmetic operators ```r a + b #addition ``` ``` ## [1] 21 ``` ```r a - b #subtraction ``` ``` ## [1] -1 ``` ```r a * b #multiplication ``` ``` ## [1] 110 ``` ```r a/b #division ``` ``` ## [1] 0.9091 ``` ```r a^b #exponentiation ``` ``` ## [1] 1e+11 ``` ```r a%/%b #integer division ``` ``` ## [1] 0 ``` ```r a%%b #modulo (remainder) b%%a makes more sense ``` ``` ## [1] 10 ``` ### Relational operators ```r a == b #is a equal to b ``` ``` ## [1] FALSE ``` ```r a > b # a is greater than b, use < for less than ``` ``` ## [1] FALSE ``` ```r a >= b # a is greater than or equal to ``` ``` ## [1] FALSE ``` ### Logical operators ```r ! #not a!=b # is a equal to be? could also do !a==b ``` ``` ## [1] FALSE ``` ```r #& #and # | #or # && #sequential and # || #sequential or ``` ### Lets create a new variable to test the logical operators ```r x <- 1:10 # note ':' creates a sequence x != a # notice how the shorter a is recycled ``` ``` ## [1] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE FALSE ``` ```r (id <- !x >= a) # notice we assigned the results to id and use () to print this ``` ``` ## [1] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE FALSE ``` ```r x > a | x < b # check if any criteria specified is TRUE ``` ``` ## [1] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE ``` ```r x > a & x < b # check if all criteria specified is TRUE ``` ``` ## [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE ``` ### Indexing ```r a[1] # get the first 'unit' of a ``` ``` ## [1] 10 ``` ```r a[2] # trying to reference something which doesn't exist, NA stands for missing value ``` ``` ## [1] NA ``` ```r x[2] # this works because x has length >=2 ``` ``` ## [1] 2 ``` ```r x[c(1, 2, 6:8)] # can use c() to get many specific elements ``` ``` ## [1] 1 2 6 7 8 ``` ```r x[id] # we can also use a logical to get our object ``` ``` ## [1] 1 2 3 4 5 6 7 8 9 ``` ### Object properties ```r str(x) #structure, x is an integer vector (one dimension) ``` ``` ## int [1:10] 1 2 3 4 5 6 7 8 9 10 ``` ```r class(id) # is logical ``` ``` ## [1] "logical" ``` ```r length(x) # length ``` ``` ## [1] 10 ``` ### Next lets load some real data to experiment with ```r data(iris) # we use a function named data to load the iris data str(iris) # structure ``` ``` ## 'data.frame': 150 obs. of 5 variables: ## $ Sepal.Length: num 5.1 4.9 4.7 4.6 5 5.4 4.6 5 4.4 4.9 ... ## $ Sepal.Width : num 3.5 3 3.2 3.1 3.6 3.9 3.4 3.4 2.9 3.1 ... ## $ Petal.Length: num 1.4 1.4 1.3 1.5 1.4 1.7 1.4 1.5 1.4 1.5 ... ## $ Petal.Width : num 0.2 0.2 0.2 0.2 0.2 0.4 0.3 0.2 0.2 0.1 ... ## $ Species : Factor w/ 3 levels "setosa","versicolor",..: 1 1 1 1 1 1 1 1 1 1 ... ``` ```r # notice we have 2 dimensions now, rows and columns now subset the object as # object[rows,columns] iris[1:5, 3:4] ``` ``` ## Petal.Length Petal.Width ## 1 1.4 0.2 ## 2 1.4 0.2 ## 3 1.3 0.2 ## 4 1.5 0.2 ## 5 1.4 0.2 ``` ```r species <- iris$Species # in data.frames and list we can also reference columns with '$' species <- iris[, "Species"] # or by name ``` ### Lets make some plots ```r plot(iris[, 1:2]) # plot the first 2 columns ``` ![plot of chunk unnamed-chunk-10](figure/unnamed-chunk-101.png) ```r help(plot) #see what other arguments plot could take plot(iris[, 1], species) ``` ![plot of chunk unnamed-chunk-10](figure/unnamed-chunk-102.png) ```r plot(iris[, 1] ~ species) # plot is different because species is factor and formula notation '~' is used ``` ![plot of chunk unnamed-chunk-10](figure/unnamed-chunk-103.png) ```r plot(iris[, 1] ~ species, col = c("red", "green", "blue")) # add a color ``` ![plot of chunk unnamed-chunk-10](figure/unnamed-chunk-104.png) ```r plot(iris[, 1] ~ species, col = c("red", "green", "blue"), ylab = colnames(iris)[1]) # and label, use function colnames() to get column names and take the first columns name ``` ![plot of chunk unnamed-chunk-10](figure/unnamed-chunk-105.png) ```r plot(iris) # because we gave the whole data frame R calls pairs() and creates a scatterplot matrix ``` ![plot of chunk unnamed-chunk-10](figure/unnamed-chunk-106.png) ```r color <- c("red", "green", "blue")[species] # we can use the factor to subset our colors to create a color for each point plot(iris, pch = 21, bg = color, main = "My Awesome Plot!") # here we give custom point shape 'pch', border 'col' and inner color 'bg' as well as a title 'main' ``` ![plot of chunk unnamed-chunk-10](figure/unnamed-chunk-107.png) ### Next lets experiment with getting summary statics ```r mean(iris) # we want to get the mean but giving the whole data.frame with the factor does not make sense? ``` ``` ## [1] NA ``` ```r mean(iris[, 1]) # this works ``` ``` ## [1] 5.843 ``` ```r mean(iris[, "Species"]) # this was the issue, it is not numeric (we could coerce 'as.numeric' but why?) ``` ``` ## [1] NA ``` ```r mean(iris[iris$Species == "setosa", 1]) # here we subset the rows to only get values for the species setosa and return the mean for column 1 for this group ``` ``` ## [1] 5.006 ``` #### Next lets get all column means for the species setosa ```r apply(iris[iris$Species == "setosa", 1:4], 2, mean) #here we use function apply to for each column use the function mean (see help(apply)) ``` ``` ## Sepal.Length Sepal.Width Petal.Length Petal.Width ## 5.006 3.428 1.462 0.246 ``` ```r # we also removed species because we know mean wont work on this and also # causes bad behaviour for other columns ``` #### Now lets get the means for each species ```r big.l <- split(iris[, 1:4], iris$Species) # create a list holding a data.frame for each level of the species (res <- lapply(big.l, apply, 2, mean)) # get means for each species and variable ``` ``` ## $setosa ## Sepal.Length Sepal.Width Petal.Length Petal.Width ## 5.006 3.428 1.462 0.246 ## ## $versicolor ## Sepal.Length Sepal.Width Petal.Length Petal.Width ## 5.936 2.770 4.260 1.326 ## ## $virginica ## Sepal.Length Sepal.Width Petal.Length Petal.Width ## 6.588 2.974 5.552 2.026 ``` ```r data.frame(res) # get results combined ``` ``` ## setosa versicolor virginica ## Sepal.Length 5.006 5.936 6.588 ## Sepal.Width 3.428 2.770 2.974 ## Petal.Length 1.462 4.260 5.552 ## Petal.Width 0.246 1.326 2.026 ``` ```r t(data.frame(res)) # transpose results ``` ``` ## Sepal.Length Sepal.Width Petal.Length Petal.Width ## setosa 5.006 3.428 1.462 0.246 ## versicolor 5.936 2.770 4.260 1.326 ## virginica 6.588 2.974 5.552 2.026 ``` #### Using this basic syntax we can exucute any custom function that we want (more to come later on R libraries to this this in an elegant and efficient way). #### Lets define a custom function to return the formatted mean and standard deviation, and then use it in the code above. ```r mean.sd <- function(x) { paste0(round(c(mean(x), sd(x)), 1), collapse = " +/- ") # our custom function } t(sapply(big.l, apply, 2, mean.sd)) ``` ``` ## Sepal.Length Sepal.Width Petal.Length Petal.Width ## setosa "5 +/- 0.4" "3.4 +/- 0.4" "1.5 +/- 0.2" "0.2 +/- 0.1" ## versicolor "5.9 +/- 0.5" "2.8 +/- 0.3" "4.3 +/- 0.5" "1.3 +/- 0.2" ## virginica "6.6 +/- 0.6" "3 +/- 0.3" "5.6 +/- 0.6" "2 +/- 0.3" ``` ================================================ FILE: Demos/R Programming/R tutorial 1.rmd ================================================ # Getting Started with R Lesson 1 ### by Dmitry Grapov ### GOALS: ### * [getting started](#basics) ### * [basic operators](#operators) ### * [plotting](#plotting) ### * [creating data summaries](#summary) ### For the following examples we will focus on the bare bones to help us learn basic concepts ### Getting started ```{r,message=FALSE,warning=FALSE} # this is a comment R doesn't interpret this 10 # works # a # remove comment and try to type a ``` R knows what ```r 10 ``` is, it is a number which has been defined in the R environment. However a has not been defined and R throws an error when trying to print it to the console. ```{r,message=FALSE,warning=FALSE} "a" # works a<-10 # assignment b<-a+1 # using a pre assigned variable ``` When we give ```r "a" ``` we define that this object is as a character vector "a" which R understands how to print to screen. We can assign ```r a ``` to something R knows about like for instance ```r 10 ```. Having done this we can now use ```r a ``` as a variable to do whatever we want. ### Arithmetic operators ```{r,message=FALSE,warning=FALSE} a + b #addition a - b #subtraction a * b #multiplication a / b #division a^b #exponentiation a %/% b #integer division a %% b #modulo (remainder) b%%a makes more sense ``` ### Relational operators ```{r,message=FALSE,warning=FALSE} a==b #is a equal to b a>b # a is greater than b, use < for less than a>=b # a is greater than or equal to ``` ### Logical operators ```{r,message=FALSE,warning=FALSE} ! #not a!=b # is a equal to be? could also do !a==b #& #and # | #or # && #sequential and # || #sequential or ``` ### Lets create a new variable to test the logical operators ```{r,message=FALSE,warning=FALSE} x<-1:10 # note ':' creates a sequence x!=a # notice how the shorter a is recycled (id<-!x>=a) # notice we assigned the results to id and use () to print this x>a|xa&x=2 x[c(1,2,6:8)] # can use c() to get many specific elements x[id] # we can also use a logical to get our object ``` ### Object properties ```{r,message=FALSE,warning=FALSE} str(x) #structure, x is an integer vector (one dimension) class(id) # is logical length(x) # length ``` ### Next lets load some real data to experiment with ```{r,message=FALSE,warning=FALSE} data(iris) # we use a function named data to load the iris data str(iris) # structure #notice we have 2 dimensions now, rows and columns # now subset the object as object[rows,columns] iris[1:5,3:4] species<-iris$Species # in data.frames and list we can also reference columns with '$' species<-iris[,"Species"] # or by name ``` ### Lets make some plots ```{r,message=FALSE,warning=FALSE} plot(iris[,1:2]) # plot the first 2 columns help(plot) #see what other arguments plot could take plot(iris[,1],species) plot(iris[,1]~species) # plot is different because species is factor and formula notation "~" is used plot(iris[,1]~species, col =c("red","green","blue")) # add a color plot(iris[,1]~species, col =c("red","green","blue"),ylab=colnames(iris)[1]) # and label, use function colnames() to get column names and take the first columns name plot(iris) # because we gave the whole data frame R calls pairs() and creates a scatterplot matrix color<-c("red","green","blue")[species] # we can use the factor to subset our colors to create a color for each point plot(iris, pch=21,bg=color, main="My Awesome Plot!") # here we give custom point shape 'pch', border 'col' and inner color 'bg' as well as a title 'main' ``` ### Next lets experiment with getting summary statics ```{r,message=FALSE,warning=FALSE} mean(iris) # we want to get the mean but giving the whole data.frame with the factor does not make sense? mean(iris[,1]) # this works mean(iris[,"Species"]) # this was the issue, it is not numeric (we could coerce 'as.numeric' but why?) mean(iris[iris$Species=="setosa",1]) # here we subset the rows to only get values for the species setosa and return the mean for column 1 for this group ``` #### Next lets get all column means for the species setosa ```{r,message=FALSE,warning=FALSE} apply(iris[iris$Species=="setosa",1:4],2,mean) #here we use function apply to for each column use the function mean (see help(apply)) # we also removed species because we know mean wont work on this and also causes bad behaviour for other columns ``` #### Now lets get the means for each species ```{r,message=FALSE,warning=FALSE} big.l<-split(iris[,1:4], iris$Species) # create a list holding a data.frame for each level of the species (res<-lapply(big.l,apply,2,mean)) # get means for each species and variable data.frame(res) # get results combined t(data.frame(res)) # transpose results ``` #### Using this basic syntax we can exucute any custom function that we want (more to come later on R libraries to this this in an elegant and efficient way). #### Lets define a custom function to return the formatted mean and standard deviation, and then use it in the code above. ```{r,message=FALSE,warning=FALSE} mean.sd<-function(x){ paste0(round(c(mean(x),sd(x)),1),collapse=" +/- ") # our custom function } t(sapply(big.l,apply,2,mean.sd)) ``` ================================================ FILE: Demos/Statistical Analysis/Covariate Adjustment.Rmd ================================================ Data Covariate Adjustment ======================================================== Covariate adjustment is a widely used approach in statistical data analysis to improve the power of tests on independent variables. In this context, covariate adjustment plays an integral role the Analysis of Covariance (ANCOVA). However the reader should be warned that the valid application of ANCOVA makes a few assuptions. For a well written discussion of this topic the reader is directed [Misunderstanding Analysis od Covariance](http://www.ncbi.nlm.nih.gov/pubmed/11261398). Covariate adjustment can also be a useful approach in data pre-processing in the context of multivariate modeling. The following is an example application of covariate adjustment using a linear model. The function [covar.adjustment](https://github.com/dgrapov/devium/blob/master/R/Devium%20Statistics.r) part of the [Devium](https://github.com/dgrapov/devium) tool set is used to carry out covariate adjustment on the famous [Iris](http://en.wikipedia.org/wiki/Iris_flower_data_set) data set. Here are the major steps involved - [loading Devium](#load) - [prepare Iris data](#prepare) - [visualize raw data](#rawvis) - [covariate adjust for Species](#covaradj) - [visualize adjusted data](#adjvis) - [conclusion](#conclusion) ### Load Devium Library ```{r,hide=FALSE,message=FALSE,warning=FALSE} source("http://pastebin.com/raw.php?i=UyDBTA57")# source Devium ``` ### Prepare Iris data ```{r} data(iris) ``` ### Visualize raw data ```{r} plot(data.frame(iris[,!colnames(iris)%in%"Species"]),pch=21, bg=rainbow(nlevels(iris$Species),alpha=.75)[iris$Species],cex=2) ``` Note the difference in the relationship between Sepal.width and Sepal.length for different species of Iris. We may want to adjust all flower measurements to model this relationship independnet of species. However be warned this may be an invalid assumption (see Introduction). ### Create Species adjusted data ```{r} factor<-iris$Species formula<-"factor" data<-iris[,!colnames(iris)%in%"Species"] adj.iris<-covar.adjustment(data,formula) ``` The adjustment is done by creating a linear model for each variable and Species labels. The "Species" adjusted data is then the residuals from this model. ### Visualize data adjusted for Species differences ```{r} plot(as.data.frame(adj.iris),pch=21, bg=rainbow(nlevels(iris$Species),alpha=.75)[iris$Species],cex=2) ``` Now all the differences in the relationships among variables due to different species is removed (which may make no sense). We can check this assumption on a multivariate basis using principal components analysis (PCA) and partial least squared discriminant analysis (O-PLS-DA). ### Principal Components Analysis (PCA) Compare the PCA scores between the raw and covariate adjusted data. ```{r} # set PCA defaults pca.inputs<-list() pca.inputs$pca.algorithm<-"svd" pca.inputs$pca.components<-2 pca.inputs$pca.center<-TRUE pca.inputs$pca.scaling<-"uv" pca.inputs$pca.cv<-"q2" #calculate raw data PCA pca.inputs$pca.data<-data pca.raw<-devium.pca.calculate(pca.inputs,return="list",plot=FALSE) #adjusted data PCA pca.inputs$pca.data<-adj.iris pca.adj<-devium.pca.calculate(pca.inputs,return="list",plot=FALSE) ``` ### Plot PCA scores for both data sets Raw data ```{r,message=FALSE,warning=FALSE} #plot.PCA(pca.raw, results = c("screeplot"),size=3,color=data.frame(species=factor), label=FALSE) plot.PCA(pca.raw, results = c("scores"),size=3,color=data.frame(species=factor), label=FALSE) ``` Adjusted data ```{r,message=FALSE,warning=FALSE} #plot.PCA(pca.adj, results = c("screeplot"),size=3,color=data.frame(species=factor), label=FALSE) plot.PCA(pca.adj, results = c("scores"),size=3,color=data.frame(species=factor), label=FALSE) ``` Comparison of the raw and adjusted scores first two principal components shows similarities in samples given all the variables. Proximity or distance between diffrent species scores can be directly translated to similarity or dissimilarity in measured values. ### O-PLS-DA A similair example can be shown using a supervised projection method, orthogonal partial least squares discriminant analysis, which attempts to maximize the diffrences between species. Next we will fit O-PLS-DA models to discriminate species using the raw and adjusted data and overview these models error for species classfication and scores. ```{r,message=FALSE,warning=FALSE} groups<-data.frame(species=factor) pls.y<-data.frame(as.numeric(groups[,1])) data<-data scaled.data1<-data.frame(scale(data,center=TRUE,scale=TRUE)) scaled.data2<-data.frame(scale(adj.iris,center=TRUE,scale=TRUE)) #make models mods1<-make.OSC.PLS.model(pls.y,pls.data=scaled.data1,comp=2,OSC.comp=1, validation = "LOO",method="oscorespls", cv.scale=TRUE,return.obj="stats",progress=FALSE) mods2<-make.OSC.PLS.model(pls.y,pls.data=scaled.data2,comp=2,OSC.comp=1, validation = "LOO",method="oscorespls", cv.scale=TRUE,return.obj="stats",progress=FALSE) ``` View the error for the classfication of raw and adjusted data. #### raw ```{r,message=FALSE,warning=FALSE} plot.OSC.results(mods1,plot="RMSEP",groups=groups) ``` #### adjusted ```{r,message=FALSE,warning=FALSE} plot.OSC.results(mods2,plot="RMSEP",groups=groups) ``` Now we see that after removing the variance in measurements due to diffrences between species our error of prediction gets very high, and is interstingly worse than what we would expect by random chance (RMSEP=0.5). Next we can compare the O-PLS-DA scores from these two models. #### raw scores ```{r,message=FALSE,warning=FALSE} final<-get.OSC.model(obj=mods1,OSC.comp=1) plot.PLS(final,results="scores",color=groups,label=FALSE) ``` #### adjusted scores ```{r,message=FALSE,warning=FALSE} final<-results<-get.OSC.model(obj=mods2,OSC.comp=1) plot.PLS(final,results="scores",color=groups,label=FALSE) ``` ### Conclusion of adjustment In this simple example of covariate adjustment using a linear model all that is really happening is the intercept is now the same for all relationships/species. We can confirm this looking at all the pairwise relationships or on a multivariate exploratory level through PCA, or using the supervised approch of O-PLS-DA. © Dmitry Grapov (2014) Creative Commons License ================================================ FILE: Demos/Statistical Analysis/Covariate Adjustment.html ================================================ Data Covariate Adjustment

Data Covariate Adjustment

Covariate adjustment is a widely used approach in statistical data analysis to improve the power of tests on independent variables. In this context, covariate adjustment plays an integral role the Analysis of Covariance (ANCOVA). However the reader should be warned that the valid application of ANCOVA makes a few assuptions. For a well written discussion of this topic the reader is directed Misunderstanding Analysis od Covariance.

Covariate adjustment can also be a useful approach in data pre-processing in the context of multivariate modeling. The following is an example application of covariate adjustment using a linear model. The function covar.adjustment part of the Devium tool set is used to carry out covariate adjustment on the famous Iris data set.

Here are the major steps involved

Load Devium Library

source("http://pastebin.com/raw.php?i=UyDBTA57")  # source Devium

Prepare Iris data

data(iris)

Visualize raw data

plot(data.frame(iris[, !colnames(iris) %in% "Species"]), pch = 21, bg = rainbow(nlevels(iris$Species), 
    alpha = 0.75)[iris$Species], cex = 2)

plot of chunk unnamed-chunk-3

Note the difference in the relationship between Sepal.width and Sepal.length for different species of Iris. We may want to adjust all flower measurements to model this relationship independnet of species. However be warned this may be an invalid assumption (see Introduction).

Create Species adjusted data

factor <- iris$Species
formula <- "factor"
data <- iris[, !colnames(iris) %in% "Species"]
adj.iris <- covar.adjustment(data, formula)

The adjustment is done by creating a linear model for each variable and Species labels. The “Species” adjusted data is then the residuals from this model.

Visualize data adjusted for Species differences

plot(as.data.frame(adj.iris), pch = 21, bg = rainbow(nlevels(iris$Species), 
    alpha = 0.75)[iris$Species], cex = 2)

plot of chunk unnamed-chunk-5

Now all the differences in the relationships among variables due to different species is removed (which may make no sense). We can check this assumption on a multivariate basis using principal components analysis (PCA) and partial least squared discriminant analysis (O-PLS-DA).

Principal Components Analysis (PCA)

Compare the PCA scores between the raw and covariate adjusted data.

# set PCA defaults
pca.inputs <- list()
pca.inputs$pca.algorithm <- "svd"
pca.inputs$pca.components <- 2
pca.inputs$pca.center <- TRUE
pca.inputs$pca.scaling <- "uv"
pca.inputs$pca.cv <- "q2"

# calculate raw data PCA
pca.inputs$pca.data <- data
pca.raw <- devium.pca.calculate(pca.inputs, return = "list", plot = FALSE)

# adjusted data PCA
pca.inputs$pca.data <- adj.iris
pca.adj <- devium.pca.calculate(pca.inputs, return = "list", plot = FALSE)

Plot PCA scores for both data sets

Raw data

# plot.PCA(pca.raw, results =
# c('screeplot'),size=3,color=data.frame(species=factor), label=FALSE)
plot.PCA(pca.raw, results = c("scores"), size = 3, color = data.frame(species = factor), 
    label = FALSE)

plot of chunk unnamed-chunk-7

Adjusted data

# plot.PCA(pca.adj, results =
# c('screeplot'),size=3,color=data.frame(species=factor), label=FALSE)
plot.PCA(pca.adj, results = c("scores"), size = 3, color = data.frame(species = factor), 
    label = FALSE)

plot of chunk unnamed-chunk-8

Comparison of the raw and adjusted scores first two principal components shows similarities in samples given all the variables. Proximity or distance between diffrent species scores can be directly translated to similarity or dissimilarity in measured values.

O-PLS-DA

A similair example can be shown using a supervised projection method, orthogonal partial least squares discriminant analysis, which attempts to maximize the diffrences between species. Next we will fit O-PLS-DA models to discriminate species using the raw and adjusted data and overview these models error for species classfication and scores.


groups <- data.frame(species = factor)
pls.y <- data.frame(as.numeric(groups[, 1]))

data <- data
scaled.data1 <- data.frame(scale(data, center = TRUE, scale = TRUE))
scaled.data2 <- data.frame(scale(adj.iris, center = TRUE, scale = TRUE))
# make models
mods1 <- make.OSC.PLS.model(pls.y, pls.data = scaled.data1, comp = 2, OSC.comp = 1, 
    validation = "LOO", method = "oscorespls", cv.scale = TRUE, return.obj = "stats", 
    progress = FALSE)
mods2 <- make.OSC.PLS.model(pls.y, pls.data = scaled.data2, comp = 2, OSC.comp = 1, 
    validation = "LOO", method = "oscorespls", cv.scale = TRUE, return.obj = "stats", 
    progress = FALSE)

View the error for the classfication of raw and adjusted data.

raw

plot.OSC.results(mods1, plot = "RMSEP", groups = groups)

plot of chunk unnamed-chunk-10

adjusted

plot.OSC.results(mods2, plot = "RMSEP", groups = groups)

plot of chunk unnamed-chunk-11

Now we see that after removing the variance in measurements due to diffrences between species our error of prediction gets very high, and is interstingly worse than what we would expect by random chance (RMSEP=0.5). Next we can compare the O-PLS-DA scores from these two models.

raw scores

final <- get.OSC.model(obj = mods1, OSC.comp = 1)
plot.PLS(final, results = "scores", color = groups, label = FALSE)

plot of chunk unnamed-chunk-12

adjusted scores

final <- results <- get.OSC.model(obj = mods2, OSC.comp = 1)
plot.PLS(final, results = "scores", color = groups, label = FALSE)

plot of chunk unnamed-chunk-13

Conclusion of adjustment

In this simple example of covariate adjustment using a linear model all that is really happening is the intercept is now the same for all relationships/species. We can confirm this looking at all the pairwise relationships or on a multivariate exploratory level through PCA, or using the supervised approch of O-PLS-DA.

© Dmitry Grapov (2014) Creative Commons License

================================================ FILE: Demos/Statistical Analysis/Covariate Adjustment.md ================================================ Data Covariate Adjustment ======================================================== Covariate adjustment is a widely used approach in statistical data analysis to improve the power of tests on independent variables. In this context, covariate adjustment plays an integral role the Analysis of Covariance (ANCOVA). However the reader should be warned that the valid application of ANCOVA makes a few assuptions. For a well written discussion of this topic the reader is directed [Misunderstanding Analysis of Covariance](http://www.ncbi.nlm.nih.gov/pubmed/11261398). Covariate adjustment can also be a useful approach in data pre-processing in the context of multivariate modeling. The following is an example application of covariate adjustment using a linear model. The function [covar.adjustment](https://github.com/dgrapov/devium/blob/master/R/Devium%20Statistics.r) part of the [Devium](https://github.com/dgrapov/devium) tool set is used to carry out covariate adjustment on the famous [Iris](http://en.wikipedia.org/wiki/Iris_flower_data_set) data set. Here are the major steps involved - [loading Devium](#load) - [prepare Iris data](#prepare) - [visualize raw data](#rawvis) - [covariate adjust for Species](#covaradj) - [visualize adjusted data](#adjvis) - [conclusion](#conclusion) ### Load Devium Library ```r source("http://pastebin.com/raw.php?i=UyDBTA57") # source Devium ``` ### Prepare Iris data ```r data(iris) ``` ### Visualize raw data ```r plot(data.frame(iris[, !colnames(iris) %in% "Species"]), pch = 21, bg = rainbow(nlevels(iris$Species), alpha = 0.75)[iris$Species], cex = 2) ``` ![plot of chunk unnamed-chunk-3](figure/unnamed-chunk-3.png) Note the difference in the relationship between Sepal.width and Sepal.length for different species of Iris. We may want to adjust all flower measurements to model this relationship independnet of species. However be warned this may be an invalid assumption (see Introduction). ### Create Species adjusted data ```r factor <- iris$Species formula <- "factor" data <- iris[, !colnames(iris) %in% "Species"] adj.iris <- covar.adjustment(data, formula) ``` The adjustment is done by creating a linear model for each variable and Species labels. The "Species" adjusted data is then the residuals from this model. ### Visualize data adjusted for Species differences ```r plot(as.data.frame(adj.iris), pch = 21, bg = rainbow(nlevels(iris$Species), alpha = 0.75)[iris$Species], cex = 2) ``` ![plot of chunk unnamed-chunk-5](figure/unnamed-chunk-5.png) Now all the differences in the relationships among variables due to different species is removed (which may make no sense). We can check this assumption on a multivariate basis using principal components analysis (PCA) and partial least squared discriminant analysis (O-PLS-DA). ### Principal Components Analysis (PCA) Compare the PCA scores between the raw and covariate adjusted data. ```r # set PCA defaults pca.inputs <- list() pca.inputs$pca.algorithm <- "svd" pca.inputs$pca.components <- 2 pca.inputs$pca.center <- TRUE pca.inputs$pca.scaling <- "uv" pca.inputs$pca.cv <- "q2" # calculate raw data PCA pca.inputs$pca.data <- data pca.raw <- devium.pca.calculate(pca.inputs, return = "list", plot = FALSE) # adjusted data PCA pca.inputs$pca.data <- adj.iris pca.adj <- devium.pca.calculate(pca.inputs, return = "list", plot = FALSE) ``` ### Plot PCA scores for both data sets Raw data ```r # plot.PCA(pca.raw, results = # c('screeplot'),size=3,color=data.frame(species=factor), label=FALSE) plot.PCA(pca.raw, results = c("scores"), size = 3, color = data.frame(species = factor), label = FALSE) ``` ![plot of chunk unnamed-chunk-7](figure/unnamed-chunk-7.png) Adjusted data ```r # plot.PCA(pca.adj, results = # c('screeplot'),size=3,color=data.frame(species=factor), label=FALSE) plot.PCA(pca.adj, results = c("scores"), size = 3, color = data.frame(species = factor), label = FALSE) ``` ![plot of chunk unnamed-chunk-8](figure/unnamed-chunk-8.png) Comparison of the raw and adjusted scores first two principal components shows similarities in samples given all the variables. Proximity or distance between diffrent species scores can be directly translated to similarity or dissimilarity in measured values. ### O-PLS-DA A similair example can be shown using a supervised projection method, orthogonal partial least squares discriminant analysis, which attempts to maximize the diffrences between species. Next we will fit O-PLS-DA models to discriminate species using the raw and adjusted data and overview these models error for species classfication and scores. ```r groups <- data.frame(species = factor) pls.y <- data.frame(as.numeric(groups[, 1])) data <- data scaled.data1 <- data.frame(scale(data, center = TRUE, scale = TRUE)) scaled.data2 <- data.frame(scale(adj.iris, center = TRUE, scale = TRUE)) # make models mods1 <- make.OSC.PLS.model(pls.y, pls.data = scaled.data1, comp = 2, OSC.comp = 1, validation = "LOO", method = "oscorespls", cv.scale = TRUE, return.obj = "stats", progress = FALSE) mods2 <- make.OSC.PLS.model(pls.y, pls.data = scaled.data2, comp = 2, OSC.comp = 1, validation = "LOO", method = "oscorespls", cv.scale = TRUE, return.obj = "stats", progress = FALSE) ``` View the error for the classfication of raw and adjusted data. #### raw ```r plot.OSC.results(mods1, plot = "RMSEP", groups = groups) ``` ![plot of chunk unnamed-chunk-10](figure/unnamed-chunk-10.png) #### adjusted ```r plot.OSC.results(mods2, plot = "RMSEP", groups = groups) ``` ![plot of chunk unnamed-chunk-11](figure/unnamed-chunk-11.png) Now we see that after removing the variance in measurements due to diffrences between species our error of prediction gets very high, and is interstingly worse than what we would expect by random chance (RMSEP=0.5). Next we can compare the O-PLS-DA scores from these two models. #### raw scores ```r final <- get.OSC.model(obj = mods1, OSC.comp = 1) plot.PLS(final, results = "scores", color = groups, label = FALSE) ``` ![plot of chunk unnamed-chunk-12](figure/unnamed-chunk-12.png) #### adjusted scores ```r final <- results <- get.OSC.model(obj = mods2, OSC.comp = 1) plot.PLS(final, results = "scores", color = groups, label = FALSE) ``` ![plot of chunk unnamed-chunk-13](figure/unnamed-chunk-13.png) ### Conclusion of adjustment In this simple example of covariate adjustment using a linear model all that is really happening is the intercept is now the same for all relationships/species. We can confirm this looking at all the pairwise relationships or on a multivariate exploratory level through PCA, or using the supervised approch of O-PLS-DA. © Dmitry Grapov (2014) Creative Commons License ================================================ FILE: Demos/Statistical Analysis/style.R ================================================ options(rstudio.markdownToHTML = function(inputFile, outputFile) { require(markdown) markdownToHTML(inputFile, outputFile, stylesheet=system.file("misc", "docco-template.html", package = "knitr")) } ) ================================================ FILE: Demos/Translating Between Chemical Identifiers/Translations.r ================================================ #load necessary functions #install background packages install.packages("devtools");install.packages("RJSONIO") library(devtools);library(RJSONIO) #install packages for translations # The Chemical Translation System install_github(repo = "CTSgetR", username = "dgrapov") library(CTSgetR) #for the Chemical Identifier Resolver (CIR) install_github(repo = "CIRgetR", username = "dgrapov") library(CIRgetR) #InchiKeys used for example id<-c("ZKHQWZAMYRWXGA-KQYNXXCUSA-N", "BAWFJGJZGIEFAR-NNYOXOHSSA-O","QNAYBMKLOCPYGJ-REOHCLBHSA-N") #create.csv to simulate loading fromm .csv (file written to current directory, getwd()) write.csv(data.frame(InchiKey=id),file="InchIKeys.csv",row.names=FALSE) #create.csv to simulate loading fromm .csv #upload .csv id<-read.csv(file="InchIKeys.csv",header=TRUE) # Goal: translate from inchiKeys to ChemSpider Ids #use Chemical Identifier Resolver (CIR) by the CADD Group at the NCI/NIH results<-CIRgetR(id,to= "chemspider_id",return.all=FALSE) #use the Chemical Translation System results2<-CTSgetR(id,from="InChIKey",to="ChemSpider",parallel=FALSE) #are there any differences between two results? miss.match<-!as.matrix(results2)%in%as.matrix(results)|!as.matrix(results)%in%as.matrix(results2) paste(sum(miss.match),"difference(s) between results",sep=" ") data.frame(CIR= results[,1], CTS = results2[,1])[miss.match,]#two different records for both are Alanine #CTS (but not CIR) can be used to generate InChI key/code from identifier CSid<-results[miss.match,] # convert CIR ChemSpider Id to inChiKey results3<-CTSgetR(CSid,from="ChemSpider",to="InChIKey",parallel=FALSE) #compare keys <<<<<<< HEAD if(as.matrix(results3)==as.matrix(id[miss.match,,drop=FALSE]))cat("codes match!","\n") else cat("codes DO NOT match!","\n") ======= if(results3==id[miss.match])cat("codes match!","\n") else cat("codes DO NOT match!","\n") >>>>>>> 527fe4c248ad2040b68cecee54e55a80e05bade5 #here is a more advanced example for translating from one ID to many ##translate InchI Key to allpossible options available in CIR CIR.options<-c("smiles", "names", "iupac_name", "cas", "inchi", "stdinchi", "inchikey", "stdinchikey", "ficts", "ficus", "uuuuu", "image", "file", "mw", "monoisotopic_mass","chemspider_id", "pubchem_sid", "chemnavigator_sid", "formula", "chemnavigator_sid") all.results.CIR<-sapply(1:length(CIR.options), function(i) { cat(CIR.options[i],"\n") CIRgetR(id=id,to=CIR.options[i],return.all=FALSE) }) names(all.results.CIR)<-CIR.options all.results.CIR<-data.frame(all.results.CIR )# object #get all possible options from CTS CTS.options<-CTS.options() CTS.options # see options id<-results2 all.results.CTS<-sapply(1:length(CTS.options), function(i) { cat(CTS.options[i],"\n") CTSgetR(id=id,to=CTS.options[i],from="ChemSpider") }) names(all.results.CTS)<-CTS.options all.results.CTS<-data.frame(all.results.CTS) # object #calculate % error for each querry as a percent of asked translations CIR.error<-round(((sum(unlist(all.results.CIR)=="

Page not found (404)

")/length(unlist(id))))/length(CIR.options)*100,0) CTS.error<-round((sum(unlist(all.results.CTS)=="error")/length(unlist(id)))/length(CTS.options)*100,0) data.frame(CIR.error=CIR.error,CTS.error=CTS.error) #choose best <<<<<<< HEAD best<-c("all.results.CIR","all.results.CTS")[which.min(c(CIR.error,CTS.error))[1]] # [1] for tie breaker cat("Best results: ",best, "\n") ======= best<-c("all.results.CIR","all.results.CTS")[which.min(c(CIR.error,CST.error))[1]] # [1] for tie breaker >>>>>>> 527fe4c248ad2040b68cecee54e55a80e05bade5 #save the best result to a .csv write.csv(get(best),file="best translation.csv") #get image for querry using CIR <<<<<<< HEAD #get image URL from InchIKey id<-"ZKHQWZAMYRWXGA-KQYNXXCUSA-N" image.url<-as.character(unlist(CIRgetR(id,to= "image",return.all=FALSE) )) download.file(image.url,"image.gif") install.packages("caTools") library(caTools) gif <- read.gif(image.url, verbose = TRUE, flip = TRUE) par(pin=c(2,2)) # change this for multiple gif results ======= download.file("http://cactus.nci.nih.gov/chemical/structure/ZKHQWZAMYRWXGA-KQYNXXCUSA-N/image","image.gif") install.packages("caTools") library(caTools) gif <- read.gif(image.url, verbose = TRUE, flip = TRUE) par(pin=c(2,2)) >>>>>>> 527fe4c248ad2040b68cecee54e55a80e05bade5 image(gif$image, col = gif$col, main = gif$comment, frame.plot=FALSE,xaxt="n", yaxt="n") ================================================ FILE: Demos/dplyr/Tutorials.Rproj ================================================ Version: 1.0 RestoreWorkspace: Default SaveWorkspace: Default AlwaysSaveHistory: Default EnableCodeIndexing: Yes UseSpacesForTab: Yes NumSpacesForTab: 2 Encoding: UTF-8 RnwWeave: Sweave LaTeX: pdfLaTeX ================================================ FILE: Demos/dplyr/hands_on_with_dplyr.Rmd ================================================ --- title: "Hands-on with dplyr" author: "Dmitry Grapov" output: html_document: keep_md: yes --- ## Introduction This is meant to be an introduction to [dplyr](http://cran.r-project.org/web/packages/dplyr/index.html) which covers dplyr basics, gets into a little bit of programming with dplyr and ends with brief mention of some gotchas and a benchmarking comparison to base for the split-apply strategy. You might also find [Rstudio](http://www.rstudio.com/)'s [Data Wrangling Cheat Sheet](http://www.rstudio.com/wp-content/uploads/2015/02/data-wrangling-cheatsheet.pdf) featuring dplyr useful (this is also where I borrowed some of the images used in this tutorial). The [dplyr](http://cran.r-project.org/web/packages/dplyr/index.html) package from [Hadley Wickham](http://had.co.nz/) is plain awesome. It features consistent and succinct syntax, is computationally fast and getting better with every release. The `dplyr` package has replaced many common more verbose R idioms which I had to previously rely upon for most common data analysis tasks. For example, many data analysis tasks involve the procedure of splitting the data set based on a grouping variable and then applying a function to each of the groups (split-apply). Lets say I want to calculate the median values for a few parameters for cars with different numbers of cylinders using the [mtcars](https://stat.ethz.ch/R-manual/R-devel/library/datasets/html/mtcars.html) data set. Set up the data for the example. ```{r,warning=FALSE,message=FALSE} #some data prep data(mtcars) data<-mtcars data$cyl<-factor(data$cyl) ``` Split-lapply-apply in `base`: ```{r,warning=FALSE,message=FALSE} #select some variable of interest vars<-c("mpg","wt","qsec") tmp.data<-data[,colnames(data)%in%vars] #split the data on the number of cylinders big.l<-split(tmp.data,data$cyl) #apply some function of interest to all columns results<-lapply(big.l, function(x) apply(x,2,median)) #bind results and add splitting info data.frame(cyl=names(results),do.call("rbind",results)) ``` Now the same process using `dplyr`: ```{r,warning=FALSE,message=FALSE} suppressPackageStartupMessages(library(dplyr)) #variables of interest vars<-c("mpg","wt","qsec") data %>% group_by(cyl) %>% select(one_of(vars)) %>% summarise_each(funs(median(.))) ``` Switching from `base` to `dplyr` for data manipulation feels a little like this: `base` ![](images/BD_wasted.gif) `dplyr`![](images/AD_wasted.gif) Each of the individual `dplyr` verbs are discussed in more detail below, but the use of `%>%` or the pipe operator is worth mentioning now. The `%>%` operator is imported from `magrittr` and for the purpose of this tutorial we can simply think of it as `then`. From the cheatsheet referenced above: ![](images/pipe.png) **** ## Overview of common dplyr functions and verbs I highly recommend that you take a look at the `dplyr` [vignetts](http://cran.r-project.org/web/packages/dplyr/index.html) for more detailed description of all of this packages capabilities. One immediate addition in `dplyr` you might notice is `tbl_df` which is a local data frame and mostly behaves like the classical `data.frame` but is more convenient for working with large data. ```{r} tbl_df(mtcars) # control the number of rows print(tbl_df(mtcars),n=5) ``` You can can make sure all columns are output to the screen using `options(dplyr.width = Inf)`. `glimpse` is another useful function which is an analogue of `str` but tries to show you more of the data. ```{r} str(mtcars) glimpse(mtcars) ``` For the purpose of this tutorial we will be mostly working with data.frames, however it should be noted that dplyr syntax abstracts away the need to specify the kind of object being manipulated and most everything we will cover can also be applied to interact with a variety of database objects. The most common `dplyr` functions also referred to as verbs are as follows (see more in the [introduction](http://cran.rstudio.com/web/packages/dplyr/vignettes/introduction.html) vignette): * filter() and slice() * arrange() * select() and rename() * mutate() and transmute() * summarise() * group_by() The following commands will be demonstrated using the [hflights](http://cran.r-project.org/web/packages/hflights/index.html) data set. ```{r} suppressPackageStartupMessages(library(hflights)) (flights <- tbl_df(hflights)) ``` Which contains `r nrow(flights)` records for `r ncol(flights)` variables for flights departing Houston airport for 2011. ### Filter Use `filter` to keep or select rows matching some criteria or condition(s). ![](images/filter.png) `base` ```{r} flights[flights$Month == 1 & flights$DayofMonth == 1, ] #could have also used subset subset(flights, Month == 1 & DayofMonth == 1) #be wary of using subset programmatically: http://stackoverflow.com/questions/9860090/in-r-why-is-better-than-subset ``` `dplyr` ```{r} #comma is the same as an ampersand (&) filter(flights, Month == 1, DayofMonth == 1) # use pipe(|) for or filter(flights, Month == 1 | DayofMonth == 1) ``` We can also include any of the following operators in filter. ![](images/logic.png) Slice is a variant of filter used to extract rows based on position. `base` ```{r} flights[1:10,] ``` `dplyr` ```{r} slice(flights, 1:10) ``` **** ## Arrange Order data based on specified columns. `base` ```{r} flights[order(flights$Month),] ``` `dplyr` ```{r} arrange(flights,Month) #decreasing order arrange(flights,desc(Month)) #break ties using more columns arrange(flights,desc(Month),DayOfWeek) ``` **** ### Select Select columns from the data. `base` ```{r} flights[,colnames(flights)%in%c("Month","DayOfWeek")] ``` `dplyr` ```{r} select(flights,Month,DayOfWeek) #select using a dynamic variable variables<-c("Month","DayOfWeek") select(flights,one_of(variables)) #remove variables select(flights,one_of(variables),-Month) ``` Select also provides many regular expression wrappers. ![](images/select.png) Use `rename` to change column names. ```{r} rename(flights,diverted=Diverted) ``` **** ### Mutate and Transmute Use theses verbs to create a new column variable which in the case of `mutate` will be added to the or created as a stand-alone variables `transmute`. ![](images/mutate_1.png) Lets calculate the wait time based on the difference between `ArrTime` and `DepTime`. `base` ```{r} #transmute like head(wait<-flights$ArrTime - flights$DepTime) #mutate like head(flights2<-cbind(flights,wait)) ``` `dplyr` ```{r} #stand alone transmute(flights,diff = ArrTime - DepTime) #added to data mutate(flights,diff = ArrTime - DepTime) ``` Many dplyr functions will let you use newly create variables in the same function which is creating the variable in the first place. ![](images/inception.png) ```{r} transmute(flights,diff = ArrTime - DepTime, ratio = ArrTime/diff, ratio2 = diff/ratio) ``` The `mutate_each` function can be used to apply a function to every column in the dataframe. Lets bin each column into quartiles using the `ntile` function. ```{r} glimpse(mutate_each(flights,funs(ntile(.,n=4)))) ``` Here is a list of some other `dplyr` convenience functions. ![](images/mutate_2.png) **** ### Summarise Carry out a function on the data frame returning a single value. ![](images/summarise_1.png) Lets calculate the range for wait times and use this as an opportunity to involve `%>%`. ```{r} transmute(flights,diff = ArrTime - DepTime) %>% summarise(.,min=min(diff,na.rm=TRUE),max=max(diff,na.rm=TRUE)) ``` Either we just discovered time travel or we should have been referencing the day and time in our calculation of the difference between arrival and departure. Lets count how many calculations we may have screwed up. ```{r} (bad<-transmute(flights,diff = ArrTime - DepTime) %>% filter(diff<0) %>% count(.)) ``` An error rate of `r round(bad/nrow(flights)*100,1)` %may or may not be acceptable. We can use `summarise_each` to apply a function to every column in the data set. Lets calculate the median and median absolute deviation for every numeric variable. ```{r} vars<-colnames(flights)[sapply(flights,is.numeric)] flights %>% select(one_of(vars)) %>% summarise_each(.,funs(median=median(.,na.rm=TRUE),mad=mad(.,na.rm=TRUE))) ``` Here are some additional functions which can be used with summarise. ![](images/summarise_2.png) ***** ##Group_by Break the data sets into groups of rows. ![](images/group_by.png) `group_by` adds the final piece of the puzzle we need to execute the split-apply strategy to our hearts content. This function becomes very powerful when combined with the previously discussed `dplyr` verbs. For example lets calculate which day of the week has the most cancellations. ```{r} flights %>% group_by(DayOfWeek) %>% select(Cancelled) %>% summarise_each(funs(canceled=sum(.,na.rm=TRUE), total=n(), percent_cancelled=round(canceled/total*100,1))) ``` We can also use `group_by` to generate groups using more than one variable. For example lets calculate the median `AirTime` times by `Distance` and `TailNum`. ```{r} (slowest<-flights %>% group_by(Distance,TailNum) %>% select(AirTime) %>% summarise_each(funs(mean(.,na.rm=TRUE)))) ``` Lets identify the 3 slowest and fastest planes based on `AirTime` for some arbitrary `Distance`. ```{r} #choose arbitrary distance tmp<-slowest %>% na.omit(.) %>% filter(.,Distance==781) %>% arrange(AirTime) #not clear why, but the results can't be bound directly tmp %>% head(.,3) tmp %>% tail(.,3) %>% arrange(desc(.)) ``` A common data analysis task might be to carry out some group-wise normalization or adjustments of the data. For example we may want to calculate the day of the week with the slowest flights, but also adjust for differences between individual planes. To do this we will start by calculating the average speed for each plane. ```{r} #calculate speed flights<-flights %>% mutate(hrs=AirTime/60, speed=Distance/hrs) (averages<-flights %>% group_by(TailNum) %>% select(.,speed) %>% summarise_each(funs(mean(.,na.rm=TRUE))) %>% rename(.,mean_speed=speed)) ``` Next lets express the overall speed for each plane as ratio to the mean plane speed. To do this we will use one of the powerful join capabilities in `dplyr`. ![](images/join_1.png) We will join with the original data set based on `TailNum` and calculate the plane-adjusted measure of speed. ```{r} right_join(flights,averages,by="TailNum") %>% mutate(norm_speed = speed / mean_speed) %>% group_by(DayOfWeek) %>% select(contains("speed")) %>% summarise_each(funs(mean(.,na.rm=TRUE))) ``` So if things worked out like we expected it looks like Saturday flights are fastest and Tuesday the slowest. ### Using dplyr programmatically Most of the examples up to this point featured using `dplyr` in interactive mode. However there are variants of nearly every verb which are best suited for use inside other functions. To see what theses are take a look at `verb_` versions of each function (e.g. `summarise_`). ### Benchmarking split-apply in base and dplyr Finally I will wrap with a relatively non-sophisticated benchmarking head-to-head comparison of `base` and `dplyr` speed for the almighty split-apply strategy. Lets set up the data. ```{r,warning=FALSE,message=FALSE} rows<-10000 cols<-100 groups<-100 samples<-rows/groups tmp.data<-data.frame(matrix(rnorm(rows),rows,cols)) tmp.data$group<-rep(1:groups,each=samples) ``` This data set has `r rows` rows, `r cols` columns and `r groups` groups with `r samples` samples each. `base` ![](images/sad_face.png) ```{r,warning=FALSE,message=FALSE} ptm <- Sys.time() #split the data on the number of cylinders big.l<-split(tmp.data,tmp.data$group) #apply some function of interest to all columns results<-lapply(big.l, function(x) apply(x,2,median)) #bind results and add splitting info results<-data.frame(group=names(results),do.call("rbind",results)) #elapsed time (bd<-Sys.time()-ptm ) ``` `dplyr` ![](images/happy_face.gif) ```{r,warning=FALSE,message=FALSE} ptm <- Sys.time() results<-tmp.data %>% group_by(group) %>% summarise_each(funs(median(.))) #elapsed time (ad<-Sys.time()-ptm ) ``` Wow I just saved `r bd - ad ` seconds of my life! Create a benchmark visualization comparing base to dplyr for differing number of groups, rows and columns. Uncomment the code in the appendix below and modify as needed to re-run the benchmark. ```{r,warning=FALSE,message=FALSE,fig.width = 10,fig.width = 10} load(file="benchmark results") #create a plot library(reshape2) library(ggplot2) tmp.data<-melt(res,id.vars=c("rows","columns","groups","samples")) %>% mutate(seconds=value*60) ggplot(tmp.data, aes(y=seconds,x=groups,group=variable,color=variable)) + geom_line() +geom_point()+ facet_grid(rows ~ columns) +scale_y_log10() ``` The plot above shows the calculation time for 10 replications in seconds (y-axis) for calculating the median of varying number of groups (x-axis), rows (y-facet) and columns (y-facet). ***** ### Gotchas 1. In `dplyr` rownames are a second class citizen and are not stored. 2. Even though nearly every `dplyr` tutorial features pipes (`%>%`) it may be easier to learn both `dplyr` and `%>%` separately. 3. Most `dplyr` functions only work on objects coercible to ~ `data.frames`. Lots of my debugging sessions start with trying to understand the data structure of objects I am passing to `dplyr`. ***** ### Appendix ```{r,warning=FALSE,message=FALSE} # # #set up functions to time # base_fun<-function(data){ # #split the data on the number of cylinders # big.l<-split(data,data$group) # # #apply some function of interest to all columns # results<-lapply(big.l, function(x) apply(x,2,median)) # # #bind results and add splitting info # data.frame(group=names(results),do.call("rbind",results)) # } # # dplyr_fun<-function(data){ # # data %>% group_by(group) %>% summarise_each(funs(median(.))) # } # # #benchmark function # benchmark_fun<-function(rows,cols,groups){ # # #set up data # samples<-floor(rows/groups) # tmp.data<-data.frame(matrix(rnorm(rows),rows,cols)) # tmp.data$group<-rep(1:groups,length.out=rows) # # #base # base.time<-system.time(replicate(10,base_fun(tmp.data))) # # #dplyr # dplyr.time<-system.time(replicate(10,dplyr_fun(tmp.data))) # # data.frame(rows=rows,columns=cols,groups=groups,samples=samples,base=signif(base.time["elapsed"]/10,3),dplyr=signif(dplyr.time["elapsed"]/10,3)) # # } # # #run benchmarks # len<-5 # groups<-seq(5, 100,length.out=len) %>% signif(.,0) # rows<-seq(100, 10000,length.out=len) %>% signif(.,0) # cols<-seq(10, 100,length.out=len) %>% signif(.,0) # # #benchmarks # results<-list() # counter<-1 # for(i in 1:length(groups)){ # .group<-groups[i] # for(j in 1:length(rows)){ # .row<-rows[j] # for(k in 1:length(cols)){ # .col<-cols[k] # results[[counter]]<- benchmark_fun(.row,.col,.group) # counter<-counter+1 # } # } # } # # # res<-do.call("rbind",results) # save(res,file="benchmark results") # #create a plot # library(reshape2) # library(ggplot2) # tmp.data<-melt(res,id.vars=c("rows","columns","groups","samples")) %>% mutate(seconds=value*60) # # ggplot(tmp.data, aes(y=seconds,x=groups,group=variable,color=variable)) + geom_line() +geom_point()+ facet_grid(columns ~ rows) +scale_y_log10() ``` ================================================ FILE: Demos/dplyr/hands_on_with_dplyr.html ================================================ Hands-on with dplyr

Introduction

This is meant to be an introduction to dplyr which covers dplyr basics, gets into a little bit of programming with dplyr and ends with brief mention of some gotchas and a benchmarking comparison to base for the split-apply strategy. You might also find Rstudio’s Data Wrangling Cheat Sheet featuring dplyr useful (this is also where I borrowed some of the images used in this tutorial).

The dplyr package from Hadley Wickham is plain awesome. It features consistent and succinct syntax, is computationally fast and getting better with every release. The dplyr package has replaced many common more verbose R idioms which I had to previously rely upon for most common data analysis tasks.

For example, many data analysis tasks involve the procedure of splitting the data set based on a grouping variable and then applying a function to each of the groups (split-apply). Lets say I want to calculate the median values for a few parameters for cars with different numbers of cylinders using the mtcars data set.

Set up the data for the example.

#some data prep
data(mtcars)
data<-mtcars
data$cyl<-factor(data$cyl)

Split-lapply-apply in base:

#select some variable of interest
vars<-c("mpg","wt","qsec")
tmp.data<-data[,colnames(data)%in%vars]

#split the data on the number of cylinders
big.l<-split(tmp.data,data$cyl)

#apply some function of interest to all columns
results<-lapply(big.l, function(x) apply(x,2,median))

#bind results and add splitting info
data.frame(cyl=names(results),do.call("rbind",results))
##   cyl  mpg    wt   qsec
## 4   4 26.0 2.200 18.900
## 6   6 19.7 3.215 18.300
## 8   8 15.2 3.755 17.175

Now the same process using dplyr:

suppressPackageStartupMessages(library(dplyr))

#variables of interest
vars<-c("mpg","wt","qsec")

data %>% group_by(cyl) %>% select(one_of(vars)) %>% summarise_each(funs(median(.)))
## Source: local data frame [3 x 4]
## 
##   cyl  mpg    wt   qsec
## 1   4 26.0 2.200 18.900
## 2   6 19.7 3.215 18.300
## 3   8 15.2 3.755 17.175

Switching from base to dplyr for data manipulation feels a little like this:

base

dplyr

Each of the individual dplyr verbs are discussed in more detail below, but the use of %>% or the pipe operator is worth mentioning now. The %>% operator is imported from magrittr and for the purpose of this tutorial we can simply think of it as then. From the cheatsheet referenced above:

Overview of common dplyr functions and verbs

I highly recommend that you take a look at the dplyr vignetts for more detailed description of all of this packages capabilities. One immediate addition in dplyr you might notice is tbl_df which is a local data frame and mostly behaves like the classical data.frame but is more convenient for working with large data.

tbl_df(mtcars)
## Source: local data frame [32 x 11]
## 
##     mpg cyl  disp  hp drat    wt  qsec vs am gear carb
## 1  21.0   6 160.0 110 3.90 2.620 16.46  0  1    4    4
## 2  21.0   6 160.0 110 3.90 2.875 17.02  0  1    4    4
## 3  22.8   4 108.0  93 3.85 2.320 18.61  1  1    4    1
## 4  21.4   6 258.0 110 3.08 3.215 19.44  1  0    3    1
## 5  18.7   8 360.0 175 3.15 3.440 17.02  0  0    3    2
## 6  18.1   6 225.0 105 2.76 3.460 20.22  1  0    3    1
## 7  14.3   8 360.0 245 3.21 3.570 15.84  0  0    3    4
## 8  24.4   4 146.7  62 3.69 3.190 20.00  1  0    4    2
## 9  22.8   4 140.8  95 3.92 3.150 22.90  1  0    4    2
## 10 19.2   6 167.6 123 3.92 3.440 18.30  1  0    4    4
## ..  ... ...   ... ...  ...   ...   ... .. ..  ...  ...
# control the number of rows
print(tbl_df(mtcars),n=5)
## Source: local data frame [32 x 11]
## 
##     mpg cyl disp  hp drat    wt  qsec vs am gear carb
## 1  21.0   6  160 110 3.90 2.620 16.46  0  1    4    4
## 2  21.0   6  160 110 3.90 2.875 17.02  0  1    4    4
## 3  22.8   4  108  93 3.85 2.320 18.61  1  1    4    1
## 4  21.4   6  258 110 3.08 3.215 19.44  1  0    3    1
## 5  18.7   8  360 175 3.15 3.440 17.02  0  0    3    2
## ..  ... ...  ... ...  ...   ...   ... .. ..  ...  ...

You can can make sure all columns are output to the screen using options(dplyr.width = Inf). glimpse is another useful function which is an analogue of str but tries to show you more of the data.

str(mtcars)
## 'data.frame':    32 obs. of  11 variables:
##  $ mpg : num  21 21 22.8 21.4 18.7 18.1 14.3 24.4 22.8 19.2 ...
##  $ cyl : num  6 6 4 6 8 6 8 4 4 6 ...
##  $ disp: num  160 160 108 258 360 ...
##  $ hp  : num  110 110 93 110 175 105 245 62 95 123 ...
##  $ drat: num  3.9 3.9 3.85 3.08 3.15 2.76 3.21 3.69 3.92 3.92 ...
##  $ wt  : num  2.62 2.88 2.32 3.21 3.44 ...
##  $ qsec: num  16.5 17 18.6 19.4 17 ...
##  $ vs  : num  0 0 1 1 0 1 0 1 1 1 ...
##  $ am  : num  1 1 1 0 0 0 0 0 0 0 ...
##  $ gear: num  4 4 4 3 3 3 3 4 4 4 ...
##  $ carb: num  4 4 1 1 2 1 4 2 2 4 ...
glimpse(mtcars)
## Observations: 32
## Variables:
## $ mpg  (dbl) 21.0, 21.0, 22.8, 21.4, 18.7, 18.1, 14.3, 24.4, 22.8, 19....
## $ cyl  (dbl) 6, 6, 4, 6, 8, 6, 8, 4, 4, 6, 6, 8, 8, 8, 8, 8, 8, 4, 4, ...
## $ disp (dbl) 160.0, 160.0, 108.0, 258.0, 360.0, 225.0, 360.0, 146.7, 1...
## $ hp   (dbl) 110, 110, 93, 110, 175, 105, 245, 62, 95, 123, 123, 180, ...
## $ drat (dbl) 3.90, 3.90, 3.85, 3.08, 3.15, 2.76, 3.21, 3.69, 3.92, 3.9...
## $ wt   (dbl) 2.620, 2.875, 2.320, 3.215, 3.440, 3.460, 3.570, 3.190, 3...
## $ qsec (dbl) 16.46, 17.02, 18.61, 19.44, 17.02, 20.22, 15.84, 20.00, 2...
## $ vs   (dbl) 0, 0, 1, 1, 0, 1, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, ...
## $ am   (dbl) 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, ...
## $ gear (dbl) 4, 4, 4, 3, 3, 3, 3, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 4, 4, ...
## $ carb (dbl) 4, 4, 1, 1, 2, 1, 4, 2, 2, 4, 4, 3, 3, 3, 4, 4, 4, 1, 2, ...

For the purpose of this tutorial we will be mostly working with data.frames, however it should be noted that dplyr syntax abstracts away the need to specify the kind of object being manipulated and most everything we will cover can also be applied to interact with a variety of database objects.

The most common dplyr functions also referred to as verbs are as follows (see more in the introduction vignette):

  • filter() and slice()
  • arrange()
  • select() and rename()
  • mutate() and transmute()
  • summarise()
  • group_by()

The following commands will be demonstrated using the hflights data set.

suppressPackageStartupMessages(library(hflights))
## Warning: package 'hflights' was built under R version 3.1.3
(flights <- tbl_df(hflights))
## Source: local data frame [227,496 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011     1          1         6    1400    1500            AA       428
## 2  2011     1          2         7    1401    1501            AA       428
## 3  2011     1          3         1    1352    1502            AA       428
## 4  2011     1          4         2    1403    1513            AA       428
## 5  2011     1          5         3    1405    1507            AA       428
## 6  2011     1          6         4    1359    1503            AA       428
## 7  2011     1          7         5    1359    1509            AA       428
## 8  2011     1          8         6    1355    1454            AA       428
## 9  2011     1          9         7    1443    1554            AA       428
## 10 2011     1         10         1    1443    1553            AA       428
## ..  ...   ...        ...       ...     ...     ...           ...       ...
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int)

Which contains 227496 records for 21 variables for flights departing Houston airport for 2011.

Filter

Use filter to keep or select rows matching some criteria or condition(s).

base

flights[flights$Month == 1 & flights$DayofMonth == 1, ]
## Source: local data frame [552 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011     1          1         6    1400    1500            AA       428
## 2  2011     1          1         6     728     840            AA       460
## 3  2011     1          1         6    1631    1736            AA      1121
## 4  2011     1          1         6    1756    2112            AA      1294
## 5  2011     1          1         6    1012    1347            AA      1700
## 6  2011     1          1         6    1211    1325            AA      1820
## 7  2011     1          1         6     557     906            AA      1994
## 8  2011     1          1         6    1824    2106            AS       731
## 9  2011     1          1         6     654    1124            B6       620
## 10 2011     1          1         6    1639    2110            B6       622
## ..  ...   ...        ...       ...     ...     ...           ...       ...
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int)
#could have also used subset
subset(flights, Month == 1 & DayofMonth == 1)
## Source: local data frame [552 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011     1          1         6    1400    1500            AA       428
## 2  2011     1          1         6     728     840            AA       460
## 3  2011     1          1         6    1631    1736            AA      1121
## 4  2011     1          1         6    1756    2112            AA      1294
## 5  2011     1          1         6    1012    1347            AA      1700
## 6  2011     1          1         6    1211    1325            AA      1820
## 7  2011     1          1         6     557     906            AA      1994
## 8  2011     1          1         6    1824    2106            AS       731
## 9  2011     1          1         6     654    1124            B6       620
## 10 2011     1          1         6    1639    2110            B6       622
## ..  ...   ...        ...       ...     ...     ...           ...       ...
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int)
#be wary of using subset programmatically: http://stackoverflow.com/questions/9860090/in-r-why-is-better-than-subset

dplyr

#comma is the same as an ampersand (&)
filter(flights, Month == 1, DayofMonth == 1)
## Source: local data frame [552 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011     1          1         6    1400    1500            AA       428
## 2  2011     1          1         6     728     840            AA       460
## 3  2011     1          1         6    1631    1736            AA      1121
## 4  2011     1          1         6    1756    2112            AA      1294
## 5  2011     1          1         6    1012    1347            AA      1700
## 6  2011     1          1         6    1211    1325            AA      1820
## 7  2011     1          1         6     557     906            AA      1994
## 8  2011     1          1         6    1824    2106            AS       731
## 9  2011     1          1         6     654    1124            B6       620
## 10 2011     1          1         6    1639    2110            B6       622
## ..  ...   ...        ...       ...     ...     ...           ...       ...
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int)
# use pipe(|) for or
filter(flights, Month == 1 | DayofMonth == 1)
## Source: local data frame [25,769 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011     1          1         6    1400    1500            AA       428
## 2  2011     1          2         7    1401    1501            AA       428
## 3  2011     1          3         1    1352    1502            AA       428
## 4  2011     1          4         2    1403    1513            AA       428
## 5  2011     1          5         3    1405    1507            AA       428
## 6  2011     1          6         4    1359    1503            AA       428
## 7  2011     1          7         5    1359    1509            AA       428
## 8  2011     1          8         6    1355    1454            AA       428
## 9  2011     1          9         7    1443    1554            AA       428
## 10 2011     1         10         1    1443    1553            AA       428
## ..  ...   ...        ...       ...     ...     ...           ...       ...
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int)

We can also include any of the following operators in filter.

Slice is a variant of filter used to extract rows based on position.

base

flights[1:10,]
## Source: local data frame [10 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011     1          1         6    1400    1500            AA       428
## 2  2011     1          2         7    1401    1501            AA       428
## 3  2011     1          3         1    1352    1502            AA       428
## 4  2011     1          4         2    1403    1513            AA       428
## 5  2011     1          5         3    1405    1507            AA       428
## 6  2011     1          6         4    1359    1503            AA       428
## 7  2011     1          7         5    1359    1509            AA       428
## 8  2011     1          8         6    1355    1454            AA       428
## 9  2011     1          9         7    1443    1554            AA       428
## 10 2011     1         10         1    1443    1553            AA       428
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int)

dplyr

slice(flights, 1:10)
## Source: local data frame [10 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011     1          1         6    1400    1500            AA       428
## 2  2011     1          2         7    1401    1501            AA       428
## 3  2011     1          3         1    1352    1502            AA       428
## 4  2011     1          4         2    1403    1513            AA       428
## 5  2011     1          5         3    1405    1507            AA       428
## 6  2011     1          6         4    1359    1503            AA       428
## 7  2011     1          7         5    1359    1509            AA       428
## 8  2011     1          8         6    1355    1454            AA       428
## 9  2011     1          9         7    1443    1554            AA       428
## 10 2011     1         10         1    1443    1553            AA       428
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int)

Arrange

Order data based on specified columns.

base

flights[order(flights$Month),]
## Source: local data frame [227,496 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011     1          1         6    1400    1500            AA       428
## 2  2011     1          2         7    1401    1501            AA       428
## 3  2011     1          3         1    1352    1502            AA       428
## 4  2011     1          4         2    1403    1513            AA       428
## 5  2011     1          5         3    1405    1507            AA       428
## 6  2011     1          6         4    1359    1503            AA       428
## 7  2011     1          7         5    1359    1509            AA       428
## 8  2011     1          8         6    1355    1454            AA       428
## 9  2011     1          9         7    1443    1554            AA       428
## 10 2011     1         10         1    1443    1553            AA       428
## ..  ...   ...        ...       ...     ...     ...           ...       ...
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int)

dplyr

arrange(flights,Month)
## Source: local data frame [227,496 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011     1          1         6    1400    1500            AA       428
## 2  2011     1          2         7    1401    1501            AA       428
## 3  2011     1          3         1    1352    1502            AA       428
## 4  2011     1          4         2    1403    1513            AA       428
## 5  2011     1          5         3    1405    1507            AA       428
## 6  2011     1          6         4    1359    1503            AA       428
## 7  2011     1          7         5    1359    1509            AA       428
## 8  2011     1          8         6    1355    1454            AA       428
## 9  2011     1          9         7    1443    1554            AA       428
## 10 2011     1         10         1    1443    1553            AA       428
## ..  ...   ...        ...       ...     ...     ...           ...       ...
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int)
#decreasing order
arrange(flights,desc(Month))
## Source: local data frame [227,496 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011    12         15         4    2113    2217            AA       426
## 2  2011    12         16         5    2004    2128            AA       426
## 3  2011    12         18         7    2007    2113            AA       426
## 4  2011    12         19         1    2108    2223            AA       426
## 5  2011    12         20         2    2008    2107            AA       426
## 6  2011    12         21         3    2025    2124            AA       426
## 7  2011    12         22         4    2021    2118            AA       426
## 8  2011    12         23         5    2015    2118            AA       426
## 9  2011    12         26         1    2013    2118            AA       426
## 10 2011    12         27         2    2007    2123            AA       426
## ..  ...   ...        ...       ...     ...     ...           ...       ...
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int)
#break ties using more columns
arrange(flights,desc(Month),DayOfWeek)
## Source: local data frame [227,496 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011    12         19         1    2108    2223            AA       426
## 2  2011    12         26         1    2013    2118            AA       426
## 3  2011    12          5         1     558     926            AA       466
## 4  2011    12         12         1     609     921            AA       466
## 5  2011    12         19         1     603     913            AA       466
## 6  2011    12         26         1     558     912            AA       466
## 7  2011    12          5         1    1206    1311            AA       865
## 8  2011    12         12         1    1339    1436            AA       865
## 9  2011    12         19         1    1203    1314            AA       865
## 10 2011    12         26         1    1200    1318            AA       865
## ..  ...   ...        ...       ...     ...     ...           ...       ...
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int)

Select

Select columns from the data.

base

flights[,colnames(flights)%in%c("Month","DayOfWeek")]
## Source: local data frame [227,496 x 2]
## 
##    Month DayOfWeek
## 1      1         6
## 2      1         7
## 3      1         1
## 4      1         2
## 5      1         3
## 6      1         4
## 7      1         5
## 8      1         6
## 9      1         7
## 10     1         1
## ..   ...       ...

dplyr

select(flights,Month,DayOfWeek)
## Source: local data frame [227,496 x 2]
## 
##    Month DayOfWeek
## 1      1         6
## 2      1         7
## 3      1         1
## 4      1         2
## 5      1         3
## 6      1         4
## 7      1         5
## 8      1         6
## 9      1         7
## 10     1         1
## ..   ...       ...
#select using a dynamic variable
variables<-c("Month","DayOfWeek")
select(flights,one_of(variables))
## Source: local data frame [227,496 x 2]
## 
##    Month DayOfWeek
## 1      1         6
## 2      1         7
## 3      1         1
## 4      1         2
## 5      1         3
## 6      1         4
## 7      1         5
## 8      1         6
## 9      1         7
## 10     1         1
## ..   ...       ...
#remove variables
select(flights,one_of(variables),-Month)
## Source: local data frame [227,496 x 1]
## 
##    DayOfWeek
## 1          6
## 2          7
## 3          1
## 4          2
## 5          3
## 6          4
## 7          5
## 8          6
## 9          7
## 10         1
## ..       ...

Select also provides many regular expression wrappers.

Use rename to change column names.

rename(flights,diverted=Diverted)
## Source: local data frame [227,496 x 21]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011     1          1         6    1400    1500            AA       428
## 2  2011     1          2         7    1401    1501            AA       428
## 3  2011     1          3         1    1352    1502            AA       428
## 4  2011     1          4         2    1403    1513            AA       428
## 5  2011     1          5         3    1405    1507            AA       428
## 6  2011     1          6         4    1359    1503            AA       428
## 7  2011     1          7         5    1359    1509            AA       428
## 8  2011     1          8         6    1355    1454            AA       428
## 9  2011     1          9         7    1443    1554            AA       428
## 10 2011     1         10         1    1443    1553            AA       428
## ..  ...   ...        ...       ...     ...     ...           ...       ...
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), diverted (int)

Mutate and Transmute

Use theses verbs to create a new column variable which in the case of mutate will be added to the or created as a stand-alone variables transmute.

Lets calculate the wait time based on the difference between ArrTime and DepTime.

base

#transmute like
head(wait<-flights$ArrTime - flights$DepTime)
## [1] 100 100 150 110 102 144
#mutate like
head(flights2<-cbind(flights,wait))
##   Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1 2011     1          1         6    1400    1500            AA       428
## 2 2011     1          2         7    1401    1501            AA       428
## 3 2011     1          3         1    1352    1502            AA       428
## 4 2011     1          4         2    1403    1513            AA       428
## 5 2011     1          5         3    1405    1507            AA       428
## 6 2011     1          6         4    1359    1503            AA       428
##   TailNum ActualElapsedTime AirTime ArrDelay DepDelay Origin Dest Distance
## 1  N576AA                60      40      -10        0    IAH  DFW      224
## 2  N557AA                60      45       -9        1    IAH  DFW      224
## 3  N541AA                70      48       -8       -8    IAH  DFW      224
## 4  N403AA                70      39        3        3    IAH  DFW      224
## 5  N492AA                62      44       -3        5    IAH  DFW      224
## 6  N262AA                64      45       -7       -1    IAH  DFW      224
##   TaxiIn TaxiOut Cancelled CancellationCode Diverted wait
## 1      7      13         0                         0  100
## 2      6       9         0                         0  100
## 3      5      17         0                         0  150
## 4      9      22         0                         0  110
## 5      9       9         0                         0  102
## 6      6      13         0                         0  144

dplyr

#stand alone
transmute(flights,diff = ArrTime - DepTime)
## Source: local data frame [227,496 x 1]
## 
##    diff
## 1   100
## 2   100
## 3   150
## 4   110
## 5   102
## 6   144
## 7   150
## 8    99
## 9   111
## 10  110
## ..  ...
#added to data
mutate(flights,diff = ArrTime - DepTime)
## Source: local data frame [227,496 x 22]
## 
##    Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum
## 1  2011     1          1         6    1400    1500            AA       428
## 2  2011     1          2         7    1401    1501            AA       428
## 3  2011     1          3         1    1352    1502            AA       428
## 4  2011     1          4         2    1403    1513            AA       428
## 5  2011     1          5         3    1405    1507            AA       428
## 6  2011     1          6         4    1359    1503            AA       428
## 7  2011     1          7         5    1359    1509            AA       428
## 8  2011     1          8         6    1355    1454            AA       428
## 9  2011     1          9         7    1443    1554            AA       428
## 10 2011     1         10         1    1443    1553            AA       428
## ..  ...   ...        ...       ...     ...     ...           ...       ...
## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime
##   (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr),
##   Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int),
##   CancellationCode (chr), Diverted (int), diff (int)

Many dplyr functions will let you use newly create variables in the same function which is creating the variable in the first place.

transmute(flights,diff = ArrTime - DepTime, ratio = ArrTime/diff, ratio2 = diff/ratio)
## Source: local data frame [227,496 x 3]
## 
##    diff    ratio    ratio2
## 1   100 15.00000  6.666667
## 2   100 15.01000  6.662225
## 3   150 10.01333 14.980027
## 4   110 13.75455  7.997356
## 5   102 14.77451  6.903782
## 6   144 10.43750 13.796407
## 7   150 10.06000 14.910537
## 8    99 14.68687  6.740715
## 9   111 14.00000  7.928571
## 10  110 14.11818  7.791372
## ..  ...      ...       ...

The mutate_each function can be used to apply a function to every column in the dataframe. Lets bin each column into quartiles using the ntile function.

glimpse(mutate_each(flights,funs(ntile(.,n=4))))
## Observations: 227496
## Variables:
## $ Year              (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,...
## $ Month             (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,...
## $ DayofMonth        (int) 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2,...
## $ DayOfWeek         (int) 3, 4, 1, 1, 2, 2, 3, 3, 4, 1, 1, 2, 2, 3, 3,...
## $ DepTime           (int) 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 2, 2, 2,...
## $ ArrTime           (int) 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,...
## $ UniqueCarrier     (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,...
## $ FlightNum         (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,...
## $ TailNum           (int) 3, 3, 3, 3, 3, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3,...
## $ ActualElapsedTime (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,...
## $ AirTime           (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,...
## $ ArrDelay          (int) 1, 1, 1, 3, 2, 2, 2, 1, 4, 4, 4, 3, 1, 2, 1,...
## $ DepDelay          (int) 2, 3, 1, 3, 3, 2, 2, 1, 4, 4, 4, 4, 2, 1, 2,...
## $ Origin            (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,...
## $ Dest              (int) 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,...
## $ Distance          (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,...
## $ TaxiIn            (int) 3, 3, 2, 4, 4, 3, 4, 3, 4, 3, 4, 1, 3, 2, 3,...
## $ TaxiOut           (int) 2, 1, 3, 4, 1, 2, 3, 2, 4, 4, 4, 2, 2, 3, 1,...
## $ Cancelled         (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,...
## $ CancellationCode  (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,...
## $ Diverted          (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,...

Here is a list of some other dplyr convenience functions.

Summarise

Carry out a function on the data frame returning a single value.

Lets calculate the range for wait times and use this as an opportunity to involve %>%.

transmute(flights,diff = ArrTime - DepTime) %>% summarise(.,min=min(diff,na.rm=TRUE),max=max(diff,na.rm=TRUE))
## Source: local data frame [1 x 2]
## 
##     min  max
## 1 -2322 1350

Either we just discovered time travel or we should have been referencing the day and time in our calculation of the difference between arrival and departure. Lets count how many calculations we may have screwed up.

(bad<-transmute(flights,diff = ArrTime - DepTime) %>% filter(diff<0) %>% count(.))
## Source: local data frame [1 x 1]
## 
##      n
## 1 2718

An error rate of 1.2 %may or may not be acceptable.

We can use summarise_each to apply a function to every column in the data set. Lets calculate the median and median absolute deviation for every numeric variable.

vars<-colnames(flights)[sapply(flights,is.numeric)]
flights %>% select(one_of(vars)) %>% summarise_each(.,funs(median=median(.,na.rm=TRUE),mad=mad(.,na.rm=TRUE)))
## Source: local data frame [1 x 32]
## 
##   Year_median Month_median DayofMonth_median DayOfWeek_median
## 1        2011            7                16                4
## Variables not shown: DepTime_median (int), ArrTime_median (dbl),
##   FlightNum_median (dbl), ActualElapsedTime_median (dbl), AirTime_median
##   (dbl), ArrDelay_median (dbl), DepDelay_median (int), Distance_median
##   (dbl), TaxiIn_median (dbl), TaxiOut_median (int), Cancelled_median
##   (dbl), Diverted_median (dbl), Year_mad (dbl), Month_mad (dbl),
##   DayofMonth_mad (dbl), DayOfWeek_mad (dbl), DepTime_mad (dbl),
##   ArrTime_mad (dbl), FlightNum_mad (dbl), ActualElapsedTime_mad (dbl),
##   AirTime_mad (dbl), ArrDelay_mad (dbl), DepDelay_mad (dbl), Distance_mad
##   (dbl), TaxiIn_mad (dbl), TaxiOut_mad (dbl), Cancelled_mad (dbl),
##   Diverted_mad (dbl)

Here are some additional functions which can be used with summarise.

Group_by

Break the data sets into groups of rows.

group_by adds the final piece of the puzzle we need to execute the split-apply strategy to our hearts content. This function becomes very powerful when combined with the previously discussed dplyr verbs. For example lets calculate which day of the week has the most cancellations.

flights %>% group_by(DayOfWeek) %>% 
  select(Cancelled)  %>% summarise_each(funs(canceled=sum(.,na.rm=TRUE),
                                             total=n(),
                                             percent_cancelled=round(canceled/total*100,1)))
## Source: local data frame [7 x 4]
## 
##   DayOfWeek canceled total percent_cancelled
## 1         1      344 34360               1.0
## 2         2      369 31649               1.2
## 3         3      396 31926               1.2
## 4         4      616 34902               1.8
## 5         5      663 34972               1.9
## 6         6      272 27629               1.0
## 7         7      313 32058               1.0

We can also use group_by to generate groups using more than one variable. For example lets calculate the median AirTime times by Distance and TailNum.

(slowest<-flights %>% group_by(Distance,TailNum) %>% 
  select(AirTime)  %>% summarise_each(funs(mean(.,na.rm=TRUE))))
## Source: local data frame [44,698 x 3]
## Groups: Distance
## 
##    Distance TailNum  AirTime
## 1        79  N14940 30.00000
## 2        79  N14943 29.00000
## 3        79  N17928 22.00000
## 4       127              NaN
## 5       127  N11106 29.00000
## 6       127  N11107 24.00000
## 7       127  N11155 26.00000
## 8       127  N11187 25.00000
## 9       127  N11189 26.00000
## 10      127  N11535 27.33333
## ..      ...     ...      ...

Lets identify the 3 slowest and fastest planes based on AirTime for some arbitrary Distance.

#choose arbitrary distance
tmp<-slowest %>% na.omit(.) %>% 
  filter(.,Distance==781) %>%
  arrange(AirTime)

#not clear why, but the results can't be bound directly
tmp %>% head(.,3)
## Source: local data frame [3 x 3]
## Groups: Distance
## 
##   Distance TailNum AirTime
## 1      781  N275WN      85
## 2      781  N453WN      87
## 3      781  N725SW      88
tmp %>% tail(.,3) %>% arrange(desc(.))
## Source: local data frame [3 x 3]
## Groups: Distance
## 
##   Distance TailNum AirTime
## 1      781  N474WN     155
## 2      781  N16646     127
## 3      781  N14653     139

A common data analysis task might be to carry out some group-wise normalization or adjustments of the data. For example we may want to calculate the day of the week with the slowest flights, but also adjust for differences between individual planes. To do this we will start by calculating the average speed for each plane.

#calculate speed
flights<-flights %>% mutate(hrs=AirTime/60, speed=Distance/hrs)

(averages<-flights %>%
  group_by(TailNum) %>% 
  select(.,speed) %>%
  summarise_each(funs(mean(.,na.rm=TRUE))) %>% 
  rename(.,mean_speed=speed))
## Source: local data frame [3,320 x 2]
## 
##    TailNum mean_speed
## 1                 NaN
## 2   N0EGMQ   462.9232
## 3   N10156   441.9362
## 4   N10575   413.5374
## 5   N11106   436.8909
## 6   N11107   437.8388
## 7   N11109   434.9298
## 8   N11113   438.9109
## 9   N11119   441.4825
## 10  N11121   436.0297
## ..     ...        ...

Next lets express the overall speed for each plane as ratio to the mean plane speed. To do this we will use one of the powerful join capabilities in dplyr.

We will join with the original data set based on TailNum and calculate the plane-adjusted measure of speed.

right_join(flights,averages,by="TailNum") %>%
  mutate(norm_speed = speed / mean_speed) %>% 
  group_by(DayOfWeek) %>% 
  select(contains("speed")) %>%
  summarise_each(funs(mean(.,na.rm=TRUE)))
## Source: local data frame [7 x 4]
## 
##   DayOfWeek    speed mean_speed norm_speed
## 1         1 419.2984   420.5830  0.9965149
## 2         2 418.2220   420.4509  0.9942130
## 3         3 419.2926   420.4236  0.9968347
## 4         4 420.3597   420.3783  0.9995784
## 5         5 419.9916   420.4688  0.9985305
## 6         6 427.3059   422.7704  1.0107539
## 7         7 423.7673   421.4198  1.0053550

So if things worked out like we expected it looks like Saturday flights are fastest and Tuesday the slowest.

Using dplyr programmatically

Most of the examples up to this point featured using dplyr in interactive mode. However there are variants of nearly every verb which are best suited for use inside other functions. To see what theses are take a look at verb_ versions of each function (e.g. summarise_).

Benchmarking split-apply in base and dplyr

Finally I will wrap with a relatively non-sophisticated benchmarking head-to-head comparison of base and dplyr speed for the almighty split-apply strategy.

Lets set up the data.

rows<-10000
cols<-100
groups<-100
samples<-rows/groups
tmp.data<-data.frame(matrix(rnorm(rows),rows,cols))
tmp.data$group<-rep(1:groups,each=samples)

This data set has 10^{4} rows, 100 columns and 100 groups with 100 samples each.

base 

ptm <- Sys.time()
#split the data on the number of cylinders
big.l<-split(tmp.data,tmp.data$group)

#apply some function of interest to all columns
results<-lapply(big.l, function(x) apply(x,2,median))

#bind results and add splitting info
results<-data.frame(group=names(results),do.call("rbind",results))

#elapsed time
(bd<-Sys.time()-ptm )
## Time difference of 0.5802019 secs

dplyr 

ptm <- Sys.time()

results<-tmp.data %>% group_by(group) %>% summarise_each(funs(median(.)))

#elapsed time
(ad<-Sys.time()-ptm )
## Time difference of 0.3492 secs

Wow I just saved 0.2310019 seconds of my life!

Create a benchmark visualization comparing base to dplyr for differing number of groups, rows and columns. Uncomment the code in the appendix below and modify as needed to re-run the benchmark.

load(file="benchmark results")

#create a plot
library(reshape2)
library(ggplot2)
tmp.data<-melt(res,id.vars=c("rows","columns","groups","samples")) %>% mutate(seconds=value*60)

ggplot(tmp.data, aes(y=seconds,x=groups,group=variable,color=variable)) + geom_line() +geom_point()+ facet_grid(rows ~ columns) +scale_y_log10()

The plot above shows the calculation time for 10 replications in seconds (y-axis) for calculating the median of varying number of groups (x-axis), rows (y-facet) and columns (y-facet).

Gotchas

  1. In dplyr rownames are a second class citizen and are not stored.
  2. Even though nearly every dplyr tutorial features pipes (%>%) it may be easier to learn both dplyr and %>% separately.
  3. Most dplyr functions only work on objects coercible to ~ data.frames. Lots of my debugging sessions start with trying to understand the data structure of objects I am passing to dplyr.

Appendix

# 
# #set up functions to time
# base_fun<-function(data){
#       #split the data on the number of cylinders
#     big.l<-split(data,data$group)
#     
#     #apply some function of interest to all columns
#     results<-lapply(big.l, function(x) apply(x,2,median))
#     
#     #bind results and add splitting info
#     data.frame(group=names(results),do.call("rbind",results))
# }
# 
# dplyr_fun<-function(data){
#   
#   data %>% group_by(group) %>% summarise_each(funs(median(.)))
# }
# 
# #benchmark function
# benchmark_fun<-function(rows,cols,groups){
#   
#   #set up data
#   samples<-floor(rows/groups)
#   tmp.data<-data.frame(matrix(rnorm(rows),rows,cols))
#   tmp.data$group<-rep(1:groups,length.out=rows)
#   
#   #base
#   base.time<-system.time(replicate(10,base_fun(tmp.data)))
#   
#   #dplyr
#   dplyr.time<-system.time(replicate(10,dplyr_fun(tmp.data)))
#   
#   data.frame(rows=rows,columns=cols,groups=groups,samples=samples,base=signif(base.time["elapsed"]/10,3),dplyr=signif(dplyr.time["elapsed"]/10,3))
#   
# }
# 
# #run benchmarks
# len<-5
# groups<-seq(5, 100,length.out=len) %>% signif(.,0)
# rows<-seq(100, 10000,length.out=len) %>% signif(.,0)
# cols<-seq(10, 100,length.out=len) %>% signif(.,0)
# 
# #benchmarks
# results<-list()
# counter<-1
# for(i in 1:length(groups)){ 
#   .group<-groups[i]
#   for(j in 1:length(rows)){
#     .row<-rows[j]
#     for(k in 1:length(cols)){
#       .col<-cols[k]
#        results[[counter]]<- benchmark_fun(.row,.col,.group)
#       counter<-counter+1
#     }
#   }  
# } 
# 
# 
# res<-do.call("rbind",results)
# save(res,file="benchmark results")
# #create a plot
# library(reshape2)
# library(ggplot2)
# tmp.data<-melt(res,id.vars=c("rows","columns","groups","samples")) %>% mutate(seconds=value*60)
# 
# ggplot(tmp.data, aes(y=seconds,x=groups,group=variable,color=variable)) + geom_line() +geom_point()+ facet_grid(columns ~ rows) +scale_y_log10()
================================================ FILE: Demos/dplyr/hands_on_with_dplyr.md ================================================ # Hands-on with dplyr Dmitry Grapov ## Introduction This is meant to be an introduction to [dplyr](http://cran.r-project.org/web/packages/dplyr/index.html) which covers dplyr basics, gets into a little bit of programming with dplyr and ends with brief mention of some gotchas and a benchmarking comparison to base for the split-apply strategy. You might also find [Rstudio](http://www.rstudio.com/)'s [Data Wrangling Cheat Sheet](http://www.rstudio.com/wp-content/uploads/2015/02/data-wrangling-cheatsheet.pdf) featuring dplyr useful (this is also where I borrowed some of the images used in this tutorial). The [dplyr](http://cran.r-project.org/web/packages/dplyr/index.html) package from [Hadley Wickham](http://had.co.nz/) is plain awesome. It features consistent and succinct syntax, is computationally fast and getting better with every release. The `dplyr` package has replaced many common more verbose R idioms which I had to previously rely upon for most common data analysis tasks. For example, many data analysis tasks involve the procedure of splitting the data set based on a grouping variable and then applying a function to each of the groups (split-apply). Lets say I want to calculate the median values for a few parameters for cars with different numbers of cylinders using the [mtcars](https://stat.ethz.ch/R-manual/R-devel/library/datasets/html/mtcars.html) data set. Set up the data for the example. ```r #some data prep data(mtcars) data<-mtcars data$cyl<-factor(data$cyl) ``` Split-lapply-apply in `base`: ```r #select some variable of interest vars<-c("mpg","wt","qsec") tmp.data<-data[,colnames(data)%in%vars] #split the data on the number of cylinders big.l<-split(tmp.data,data$cyl) #apply some function of interest to all columns results<-lapply(big.l, function(x) apply(x,2,median)) #bind results and add splitting info data.frame(cyl=names(results),do.call("rbind",results)) ``` ``` ## cyl mpg wt qsec ## 4 4 26.0 2.200 18.900 ## 6 6 19.7 3.215 18.300 ## 8 8 15.2 3.755 17.175 ``` Now the same process using `dplyr`: ```r suppressPackageStartupMessages(library(dplyr)) #variables of interest vars<-c("mpg","wt","qsec") data %>% group_by(cyl) %>% select(one_of(vars)) %>% summarise_each(funs(median(.))) ``` ``` ## Source: local data frame [3 x 4] ## ## cyl mpg wt qsec ## 1 4 26.0 2.200 18.900 ## 2 6 19.7 3.215 18.300 ## 3 8 15.2 3.755 17.175 ``` Switching from `base` to `dplyr` for data manipulation feels a little like this: `base` ![](images/BD_wasted.gif) `dplyr` ![](images/AD_wasted.gif) Each of the individual `dplyr` verbs are discussed in more detail below, but the use of `%>%` or the pipe operator is worth mentioning now. The `%>%` operator is imported from `magrittr` and for the purpose of this tutorial we can simply think of it as `then`. From the cheatsheet referenced above: ![](images/pipe.png) **** ## Overview of common dplyr functions and verbs I highly recommend that you take a look at the `dplyr` [vignetts](http://cran.r-project.org/web/packages/dplyr/index.html) for more detailed description of all of this packages capabilities. One immediate addition in `dplyr` you might notice is `tbl_df` which is a local data frame and mostly behaves like the classical `data.frame` but is more convenient for working with large data. ```r tbl_df(mtcars) ``` ``` ## Source: local data frame [32 x 11] ## ## mpg cyl disp hp drat wt qsec vs am gear carb ## 1 21.0 6 160.0 110 3.90 2.620 16.46 0 1 4 4 ## 2 21.0 6 160.0 110 3.90 2.875 17.02 0 1 4 4 ## 3 22.8 4 108.0 93 3.85 2.320 18.61 1 1 4 1 ## 4 21.4 6 258.0 110 3.08 3.215 19.44 1 0 3 1 ## 5 18.7 8 360.0 175 3.15 3.440 17.02 0 0 3 2 ## 6 18.1 6 225.0 105 2.76 3.460 20.22 1 0 3 1 ## 7 14.3 8 360.0 245 3.21 3.570 15.84 0 0 3 4 ## 8 24.4 4 146.7 62 3.69 3.190 20.00 1 0 4 2 ## 9 22.8 4 140.8 95 3.92 3.150 22.90 1 0 4 2 ## 10 19.2 6 167.6 123 3.92 3.440 18.30 1 0 4 4 ## .. ... ... ... ... ... ... ... .. .. ... ... ``` ```r # control the number of rows print(tbl_df(mtcars),n=5) ``` ``` ## Source: local data frame [32 x 11] ## ## mpg cyl disp hp drat wt qsec vs am gear carb ## 1 21.0 6 160 110 3.90 2.620 16.46 0 1 4 4 ## 2 21.0 6 160 110 3.90 2.875 17.02 0 1 4 4 ## 3 22.8 4 108 93 3.85 2.320 18.61 1 1 4 1 ## 4 21.4 6 258 110 3.08 3.215 19.44 1 0 3 1 ## 5 18.7 8 360 175 3.15 3.440 17.02 0 0 3 2 ## .. ... ... ... ... ... ... ... .. .. ... ... ``` You can can make sure all columns are output to the screen using `options(dplyr.width = Inf)`. `glimpse` is another useful function which is an analogue of `str` but tries to show you more of the data. ```r str(mtcars) ``` ``` ## 'data.frame': 32 obs. of 11 variables: ## $ mpg : num 21 21 22.8 21.4 18.7 18.1 14.3 24.4 22.8 19.2 ... ## $ cyl : num 6 6 4 6 8 6 8 4 4 6 ... ## $ disp: num 160 160 108 258 360 ... ## $ hp : num 110 110 93 110 175 105 245 62 95 123 ... ## $ drat: num 3.9 3.9 3.85 3.08 3.15 2.76 3.21 3.69 3.92 3.92 ... ## $ wt : num 2.62 2.88 2.32 3.21 3.44 ... ## $ qsec: num 16.5 17 18.6 19.4 17 ... ## $ vs : num 0 0 1 1 0 1 0 1 1 1 ... ## $ am : num 1 1 1 0 0 0 0 0 0 0 ... ## $ gear: num 4 4 4 3 3 3 3 4 4 4 ... ## $ carb: num 4 4 1 1 2 1 4 2 2 4 ... ``` ```r glimpse(mtcars) ``` ``` ## Observations: 32 ## Variables: ## $ mpg (dbl) 21.0, 21.0, 22.8, 21.4, 18.7, 18.1, 14.3, 24.4, 22.8, 19.... ## $ cyl (dbl) 6, 6, 4, 6, 8, 6, 8, 4, 4, 6, 6, 8, 8, 8, 8, 8, 8, 4, 4, ... ## $ disp (dbl) 160.0, 160.0, 108.0, 258.0, 360.0, 225.0, 360.0, 146.7, 1... ## $ hp (dbl) 110, 110, 93, 110, 175, 105, 245, 62, 95, 123, 123, 180, ... ## $ drat (dbl) 3.90, 3.90, 3.85, 3.08, 3.15, 2.76, 3.21, 3.69, 3.92, 3.9... ## $ wt (dbl) 2.620, 2.875, 2.320, 3.215, 3.440, 3.460, 3.570, 3.190, 3... ## $ qsec (dbl) 16.46, 17.02, 18.61, 19.44, 17.02, 20.22, 15.84, 20.00, 2... ## $ vs (dbl) 0, 0, 1, 1, 0, 1, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, ... ## $ am (dbl) 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, ... ## $ gear (dbl) 4, 4, 4, 3, 3, 3, 3, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 4, 4, ... ## $ carb (dbl) 4, 4, 1, 1, 2, 1, 4, 2, 2, 4, 4, 3, 3, 3, 4, 4, 4, 1, 2, ... ``` For the purpose of this tutorial we will be mostly working with data.frames, however it should be noted that dplyr syntax abstracts away the need to specify the kind of object being manipulated and most everything we will cover can also be applied to interact with a variety of database objects. The most common `dplyr` functions also referred to as verbs are as follows (see more in the [introduction](http://cran.rstudio.com/web/packages/dplyr/vignettes/introduction.html) vignette): * filter() and slice() * arrange() * select() and rename() * mutate() and transmute() * summarise() * group_by() The following commands will be demonstrated using the [hflights](http://cran.r-project.org/web/packages/hflights/index.html) data set. ```r suppressPackageStartupMessages(library(hflights)) ``` ``` ## Warning: package 'hflights' was built under R version 3.1.3 ``` ```r (flights <- tbl_df(hflights)) ``` ``` ## Source: local data frame [227,496 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 2 7 1401 1501 AA 428 ## 3 2011 1 3 1 1352 1502 AA 428 ## 4 2011 1 4 2 1403 1513 AA 428 ## 5 2011 1 5 3 1405 1507 AA 428 ## 6 2011 1 6 4 1359 1503 AA 428 ## 7 2011 1 7 5 1359 1509 AA 428 ## 8 2011 1 8 6 1355 1454 AA 428 ## 9 2011 1 9 7 1443 1554 AA 428 ## 10 2011 1 10 1 1443 1553 AA 428 ## .. ... ... ... ... ... ... ... ... ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int) ``` Which contains 227496 records for 21 variables for flights departing Houston airport for 2011. ### Filter Use `filter` to keep or select rows matching some criteria or condition(s). ![](images/filter.png) `base` ```r flights[flights$Month == 1 & flights$DayofMonth == 1, ] ``` ``` ## Source: local data frame [552 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 1 6 728 840 AA 460 ## 3 2011 1 1 6 1631 1736 AA 1121 ## 4 2011 1 1 6 1756 2112 AA 1294 ## 5 2011 1 1 6 1012 1347 AA 1700 ## 6 2011 1 1 6 1211 1325 AA 1820 ## 7 2011 1 1 6 557 906 AA 1994 ## 8 2011 1 1 6 1824 2106 AS 731 ## 9 2011 1 1 6 654 1124 B6 620 ## 10 2011 1 1 6 1639 2110 B6 622 ## .. ... ... ... ... ... ... ... ... ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int) ``` ```r #could have also used subset subset(flights, Month == 1 & DayofMonth == 1) ``` ``` ## Source: local data frame [552 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 1 6 728 840 AA 460 ## 3 2011 1 1 6 1631 1736 AA 1121 ## 4 2011 1 1 6 1756 2112 AA 1294 ## 5 2011 1 1 6 1012 1347 AA 1700 ## 6 2011 1 1 6 1211 1325 AA 1820 ## 7 2011 1 1 6 557 906 AA 1994 ## 8 2011 1 1 6 1824 2106 AS 731 ## 9 2011 1 1 6 654 1124 B6 620 ## 10 2011 1 1 6 1639 2110 B6 622 ## .. ... ... ... ... ... ... ... ... ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int) ``` ```r #be wary of using subset programmatically: http://stackoverflow.com/questions/9860090/in-r-why-is-better-than-subset ``` `dplyr` ```r #comma is the same as an ampersand (&) filter(flights, Month == 1, DayofMonth == 1) ``` ``` ## Source: local data frame [552 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 1 6 728 840 AA 460 ## 3 2011 1 1 6 1631 1736 AA 1121 ## 4 2011 1 1 6 1756 2112 AA 1294 ## 5 2011 1 1 6 1012 1347 AA 1700 ## 6 2011 1 1 6 1211 1325 AA 1820 ## 7 2011 1 1 6 557 906 AA 1994 ## 8 2011 1 1 6 1824 2106 AS 731 ## 9 2011 1 1 6 654 1124 B6 620 ## 10 2011 1 1 6 1639 2110 B6 622 ## .. ... ... ... ... ... ... ... ... ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int) ``` ```r # use pipe(|) for or filter(flights, Month == 1 | DayofMonth == 1) ``` ``` ## Source: local data frame [25,769 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 2 7 1401 1501 AA 428 ## 3 2011 1 3 1 1352 1502 AA 428 ## 4 2011 1 4 2 1403 1513 AA 428 ## 5 2011 1 5 3 1405 1507 AA 428 ## 6 2011 1 6 4 1359 1503 AA 428 ## 7 2011 1 7 5 1359 1509 AA 428 ## 8 2011 1 8 6 1355 1454 AA 428 ## 9 2011 1 9 7 1443 1554 AA 428 ## 10 2011 1 10 1 1443 1553 AA 428 ## .. ... ... ... ... ... ... ... ... ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int) ``` We can also include any of the following operators in filter. ![](images/logic.png) Slice is a variant of filter used to extract rows based on position. `base` ```r flights[1:10,] ``` ``` ## Source: local data frame [10 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 2 7 1401 1501 AA 428 ## 3 2011 1 3 1 1352 1502 AA 428 ## 4 2011 1 4 2 1403 1513 AA 428 ## 5 2011 1 5 3 1405 1507 AA 428 ## 6 2011 1 6 4 1359 1503 AA 428 ## 7 2011 1 7 5 1359 1509 AA 428 ## 8 2011 1 8 6 1355 1454 AA 428 ## 9 2011 1 9 7 1443 1554 AA 428 ## 10 2011 1 10 1 1443 1553 AA 428 ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int) ``` `dplyr` ```r slice(flights, 1:10) ``` ``` ## Source: local data frame [10 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 2 7 1401 1501 AA 428 ## 3 2011 1 3 1 1352 1502 AA 428 ## 4 2011 1 4 2 1403 1513 AA 428 ## 5 2011 1 5 3 1405 1507 AA 428 ## 6 2011 1 6 4 1359 1503 AA 428 ## 7 2011 1 7 5 1359 1509 AA 428 ## 8 2011 1 8 6 1355 1454 AA 428 ## 9 2011 1 9 7 1443 1554 AA 428 ## 10 2011 1 10 1 1443 1553 AA 428 ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int) ``` **** ## Arrange Order data based on specified columns. `base` ```r flights[order(flights$Month),] ``` ``` ## Source: local data frame [227,496 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 2 7 1401 1501 AA 428 ## 3 2011 1 3 1 1352 1502 AA 428 ## 4 2011 1 4 2 1403 1513 AA 428 ## 5 2011 1 5 3 1405 1507 AA 428 ## 6 2011 1 6 4 1359 1503 AA 428 ## 7 2011 1 7 5 1359 1509 AA 428 ## 8 2011 1 8 6 1355 1454 AA 428 ## 9 2011 1 9 7 1443 1554 AA 428 ## 10 2011 1 10 1 1443 1553 AA 428 ## .. ... ... ... ... ... ... ... ... ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int) ``` `dplyr` ```r arrange(flights,Month) ``` ``` ## Source: local data frame [227,496 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 2 7 1401 1501 AA 428 ## 3 2011 1 3 1 1352 1502 AA 428 ## 4 2011 1 4 2 1403 1513 AA 428 ## 5 2011 1 5 3 1405 1507 AA 428 ## 6 2011 1 6 4 1359 1503 AA 428 ## 7 2011 1 7 5 1359 1509 AA 428 ## 8 2011 1 8 6 1355 1454 AA 428 ## 9 2011 1 9 7 1443 1554 AA 428 ## 10 2011 1 10 1 1443 1553 AA 428 ## .. ... ... ... ... ... ... ... ... ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int) ``` ```r #decreasing order arrange(flights,desc(Month)) ``` ``` ## Source: local data frame [227,496 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 12 15 4 2113 2217 AA 426 ## 2 2011 12 16 5 2004 2128 AA 426 ## 3 2011 12 18 7 2007 2113 AA 426 ## 4 2011 12 19 1 2108 2223 AA 426 ## 5 2011 12 20 2 2008 2107 AA 426 ## 6 2011 12 21 3 2025 2124 AA 426 ## 7 2011 12 22 4 2021 2118 AA 426 ## 8 2011 12 23 5 2015 2118 AA 426 ## 9 2011 12 26 1 2013 2118 AA 426 ## 10 2011 12 27 2 2007 2123 AA 426 ## .. ... ... ... ... ... ... ... ... ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int) ``` ```r #break ties using more columns arrange(flights,desc(Month),DayOfWeek) ``` ``` ## Source: local data frame [227,496 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 12 19 1 2108 2223 AA 426 ## 2 2011 12 26 1 2013 2118 AA 426 ## 3 2011 12 5 1 558 926 AA 466 ## 4 2011 12 12 1 609 921 AA 466 ## 5 2011 12 19 1 603 913 AA 466 ## 6 2011 12 26 1 558 912 AA 466 ## 7 2011 12 5 1 1206 1311 AA 865 ## 8 2011 12 12 1 1339 1436 AA 865 ## 9 2011 12 19 1 1203 1314 AA 865 ## 10 2011 12 26 1 1200 1318 AA 865 ## .. ... ... ... ... ... ... ... ... ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int) ``` **** ### Select Select columns from the data. `base` ```r flights[,colnames(flights)%in%c("Month","DayOfWeek")] ``` ``` ## Source: local data frame [227,496 x 2] ## ## Month DayOfWeek ## 1 1 6 ## 2 1 7 ## 3 1 1 ## 4 1 2 ## 5 1 3 ## 6 1 4 ## 7 1 5 ## 8 1 6 ## 9 1 7 ## 10 1 1 ## .. ... ... ``` `dplyr` ```r select(flights,Month,DayOfWeek) ``` ``` ## Source: local data frame [227,496 x 2] ## ## Month DayOfWeek ## 1 1 6 ## 2 1 7 ## 3 1 1 ## 4 1 2 ## 5 1 3 ## 6 1 4 ## 7 1 5 ## 8 1 6 ## 9 1 7 ## 10 1 1 ## .. ... ... ``` ```r #select using a dynamic variable variables<-c("Month","DayOfWeek") select(flights,one_of(variables)) ``` ``` ## Source: local data frame [227,496 x 2] ## ## Month DayOfWeek ## 1 1 6 ## 2 1 7 ## 3 1 1 ## 4 1 2 ## 5 1 3 ## 6 1 4 ## 7 1 5 ## 8 1 6 ## 9 1 7 ## 10 1 1 ## .. ... ... ``` ```r #remove variables select(flights,one_of(variables),-Month) ``` ``` ## Source: local data frame [227,496 x 1] ## ## DayOfWeek ## 1 6 ## 2 7 ## 3 1 ## 4 2 ## 5 3 ## 6 4 ## 7 5 ## 8 6 ## 9 7 ## 10 1 ## .. ... ``` Select also provides many regular expression wrappers. ![](images/select.png) Use `rename` to change column names. ```r rename(flights,diverted=Diverted) ``` ``` ## Source: local data frame [227,496 x 21] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 2 7 1401 1501 AA 428 ## 3 2011 1 3 1 1352 1502 AA 428 ## 4 2011 1 4 2 1403 1513 AA 428 ## 5 2011 1 5 3 1405 1507 AA 428 ## 6 2011 1 6 4 1359 1503 AA 428 ## 7 2011 1 7 5 1359 1509 AA 428 ## 8 2011 1 8 6 1355 1454 AA 428 ## 9 2011 1 9 7 1443 1554 AA 428 ## 10 2011 1 10 1 1443 1553 AA 428 ## .. ... ... ... ... ... ... ... ... ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), diverted (int) ``` **** ### Mutate and Transmute Use theses verbs to create a new column variable which in the case of `mutate` will be added to the or created as a stand-alone variables `transmute`. ![](images/mutate_1.png) Lets calculate the wait time based on the difference between `ArrTime` and `DepTime`. `base` ```r #transmute like head(wait<-flights$ArrTime - flights$DepTime) ``` ``` ## [1] 100 100 150 110 102 144 ``` ```r #mutate like head(flights2<-cbind(flights,wait)) ``` ``` ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 2 7 1401 1501 AA 428 ## 3 2011 1 3 1 1352 1502 AA 428 ## 4 2011 1 4 2 1403 1513 AA 428 ## 5 2011 1 5 3 1405 1507 AA 428 ## 6 2011 1 6 4 1359 1503 AA 428 ## TailNum ActualElapsedTime AirTime ArrDelay DepDelay Origin Dest Distance ## 1 N576AA 60 40 -10 0 IAH DFW 224 ## 2 N557AA 60 45 -9 1 IAH DFW 224 ## 3 N541AA 70 48 -8 -8 IAH DFW 224 ## 4 N403AA 70 39 3 3 IAH DFW 224 ## 5 N492AA 62 44 -3 5 IAH DFW 224 ## 6 N262AA 64 45 -7 -1 IAH DFW 224 ## TaxiIn TaxiOut Cancelled CancellationCode Diverted wait ## 1 7 13 0 0 100 ## 2 6 9 0 0 100 ## 3 5 17 0 0 150 ## 4 9 22 0 0 110 ## 5 9 9 0 0 102 ## 6 6 13 0 0 144 ``` `dplyr` ```r #stand alone transmute(flights,diff = ArrTime - DepTime) ``` ``` ## Source: local data frame [227,496 x 1] ## ## diff ## 1 100 ## 2 100 ## 3 150 ## 4 110 ## 5 102 ## 6 144 ## 7 150 ## 8 99 ## 9 111 ## 10 110 ## .. ... ``` ```r #added to data mutate(flights,diff = ArrTime - DepTime) ``` ``` ## Source: local data frame [227,496 x 22] ## ## Year Month DayofMonth DayOfWeek DepTime ArrTime UniqueCarrier FlightNum ## 1 2011 1 1 6 1400 1500 AA 428 ## 2 2011 1 2 7 1401 1501 AA 428 ## 3 2011 1 3 1 1352 1502 AA 428 ## 4 2011 1 4 2 1403 1513 AA 428 ## 5 2011 1 5 3 1405 1507 AA 428 ## 6 2011 1 6 4 1359 1503 AA 428 ## 7 2011 1 7 5 1359 1509 AA 428 ## 8 2011 1 8 6 1355 1454 AA 428 ## 9 2011 1 9 7 1443 1554 AA 428 ## 10 2011 1 10 1 1443 1553 AA 428 ## .. ... ... ... ... ... ... ... ... ## Variables not shown: TailNum (chr), ActualElapsedTime (int), AirTime ## (int), ArrDelay (int), DepDelay (int), Origin (chr), Dest (chr), ## Distance (int), TaxiIn (int), TaxiOut (int), Cancelled (int), ## CancellationCode (chr), Diverted (int), diff (int) ``` Many dplyr functions will let you use newly create variables in the same function which is creating the variable in the first place. ![](images/inception.png) ```r transmute(flights,diff = ArrTime - DepTime, ratio = ArrTime/diff, ratio2 = diff/ratio) ``` ``` ## Source: local data frame [227,496 x 3] ## ## diff ratio ratio2 ## 1 100 15.00000 6.666667 ## 2 100 15.01000 6.662225 ## 3 150 10.01333 14.980027 ## 4 110 13.75455 7.997356 ## 5 102 14.77451 6.903782 ## 6 144 10.43750 13.796407 ## 7 150 10.06000 14.910537 ## 8 99 14.68687 6.740715 ## 9 111 14.00000 7.928571 ## 10 110 14.11818 7.791372 ## .. ... ... ... ``` The `mutate_each` function can be used to apply a function to every column in the dataframe. Lets bin each column into quartiles using the `ntile` function. ```r glimpse(mutate_each(flights,funs(ntile(.,n=4)))) ``` ``` ## Observations: 227496 ## Variables: ## $ Year (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ Month (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ DayofMonth (int) 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2,... ## $ DayOfWeek (int) 3, 4, 1, 1, 2, 2, 3, 3, 4, 1, 1, 2, 2, 3, 3,... ## $ DepTime (int) 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 2, 2, 2,... ## $ ArrTime (int) 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,... ## $ UniqueCarrier (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ FlightNum (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ TailNum (int) 3, 3, 3, 3, 3, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3,... ## $ ActualElapsedTime (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ AirTime (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ ArrDelay (int) 1, 1, 1, 3, 2, 2, 2, 1, 4, 4, 4, 3, 1, 2, 1,... ## $ DepDelay (int) 2, 3, 1, 3, 3, 2, 2, 1, 4, 4, 4, 4, 2, 1, 2,... ## $ Origin (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ Dest (int) 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,... ## $ Distance (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ TaxiIn (int) 3, 3, 2, 4, 4, 3, 4, 3, 4, 3, 4, 1, 3, 2, 3,... ## $ TaxiOut (int) 2, 1, 3, 4, 1, 2, 3, 2, 4, 4, 4, 2, 2, 3, 1,... ## $ Cancelled (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ CancellationCode (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ## $ Diverted (int) 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,... ``` Here is a list of some other `dplyr` convenience functions. ![](images/mutate_2.png) **** ### Summarise Carry out a function on the data frame returning a single value. ![](images/summarise_1.png) Lets calculate the range for wait times and use this as an opportunity to involve `%>%`. ```r transmute(flights,diff = ArrTime - DepTime) %>% summarise(.,min=min(diff,na.rm=TRUE),max=max(diff,na.rm=TRUE)) ``` ``` ## Source: local data frame [1 x 2] ## ## min max ## 1 -2322 1350 ``` Either we just discovered time travel or we should have been referencing the day and time in our calculation of the difference between arrival and departure. Lets count how many calculations we may have screwed up. ```r (bad<-transmute(flights,diff = ArrTime - DepTime) %>% filter(diff<0) %>% count(.)) ``` ``` ## Source: local data frame [1 x 1] ## ## n ## 1 2718 ``` An error rate of 1.2 %may or may not be acceptable. We can use `summarise_each` to apply a function to every column in the data set. Lets calculate the median and median absolute deviation for every numeric variable. ```r vars<-colnames(flights)[sapply(flights,is.numeric)] flights %>% select(one_of(vars)) %>% summarise_each(.,funs(median=median(.,na.rm=TRUE),mad=mad(.,na.rm=TRUE))) ``` ``` ## Source: local data frame [1 x 32] ## ## Year_median Month_median DayofMonth_median DayOfWeek_median ## 1 2011 7 16 4 ## Variables not shown: DepTime_median (int), ArrTime_median (dbl), ## FlightNum_median (dbl), ActualElapsedTime_median (dbl), AirTime_median ## (dbl), ArrDelay_median (dbl), DepDelay_median (int), Distance_median ## (dbl), TaxiIn_median (dbl), TaxiOut_median (int), Cancelled_median ## (dbl), Diverted_median (dbl), Year_mad (dbl), Month_mad (dbl), ## DayofMonth_mad (dbl), DayOfWeek_mad (dbl), DepTime_mad (dbl), ## ArrTime_mad (dbl), FlightNum_mad (dbl), ActualElapsedTime_mad (dbl), ## AirTime_mad (dbl), ArrDelay_mad (dbl), DepDelay_mad (dbl), Distance_mad ## (dbl), TaxiIn_mad (dbl), TaxiOut_mad (dbl), Cancelled_mad (dbl), ## Diverted_mad (dbl) ``` Here are some additional functions which can be used with summarise. ![](images/summarise_2.png) ***** ## Group_by Break the data sets into groups of rows. ![](images/group_by.png) `group_by` adds the final piece of the puzzle we need to execute the split-apply strategy to our hearts content. This function becomes very powerful when combined with the previously discussed `dplyr` verbs. For example lets calculate which day of the week has the most cancellations. ```r flights %>% group_by(DayOfWeek) %>% select(Cancelled) %>% summarise_each(funs(canceled=sum(.,na.rm=TRUE), total=n(), percent_cancelled=round(canceled/total*100,1))) ``` ``` ## Source: local data frame [7 x 4] ## ## DayOfWeek canceled total percent_cancelled ## 1 1 344 34360 1.0 ## 2 2 369 31649 1.2 ## 3 3 396 31926 1.2 ## 4 4 616 34902 1.8 ## 5 5 663 34972 1.9 ## 6 6 272 27629 1.0 ## 7 7 313 32058 1.0 ``` We can also use `group_by` to generate groups using more than one variable. For example lets calculate the median `AirTime` times by `Distance` and `TailNum`. ```r (slowest<-flights %>% group_by(Distance,TailNum) %>% select(AirTime) %>% summarise_each(funs(mean(.,na.rm=TRUE)))) ``` ``` ## Source: local data frame [44,698 x 3] ## Groups: Distance ## ## Distance TailNum AirTime ## 1 79 N14940 30.00000 ## 2 79 N14943 29.00000 ## 3 79 N17928 22.00000 ## 4 127 NaN ## 5 127 N11106 29.00000 ## 6 127 N11107 24.00000 ## 7 127 N11155 26.00000 ## 8 127 N11187 25.00000 ## 9 127 N11189 26.00000 ## 10 127 N11535 27.33333 ## .. ... ... ... ``` Lets identify the 3 slowest and fastest planes based on `AirTime` for some arbitrary `Distance`. ```r #choose arbitrary distance tmp<-slowest %>% na.omit(.) %>% filter(.,Distance==781) %>% arrange(AirTime) #not clear why, but the results can't be bound directly tmp %>% head(.,3) ``` ``` ## Source: local data frame [3 x 3] ## Groups: Distance ## ## Distance TailNum AirTime ## 1 781 N275WN 85 ## 2 781 N453WN 87 ## 3 781 N725SW 88 ``` ```r tmp %>% tail(.,3) %>% arrange(desc(.)) ``` ``` ## Source: local data frame [3 x 3] ## Groups: Distance ## ## Distance TailNum AirTime ## 1 781 N474WN 155 ## 2 781 N16646 127 ## 3 781 N14653 139 ``` A common data analysis task might be to carry out some group-wise normalization or adjustments of the data. For example we may want to calculate the day of the week with the slowest flights, but also adjust for differences between individual planes. To do this we will start by calculating the average speed for each plane. ```r #calculate speed flights<-flights %>% mutate(hrs=AirTime/60, speed=Distance/hrs) (averages<-flights %>% group_by(TailNum) %>% select(.,speed) %>% summarise_each(funs(mean(.,na.rm=TRUE))) %>% rename(.,mean_speed=speed)) ``` ``` ## Source: local data frame [3,320 x 2] ## ## TailNum mean_speed ## 1 NaN ## 2 N0EGMQ 462.9232 ## 3 N10156 441.9362 ## 4 N10575 413.5374 ## 5 N11106 436.8909 ## 6 N11107 437.8388 ## 7 N11109 434.9298 ## 8 N11113 438.9109 ## 9 N11119 441.4825 ## 10 N11121 436.0297 ## .. ... ... ``` Next lets express the overall speed for each plane as ratio to the mean plane speed. To do this we will use one of the powerful join capabilities in `dplyr`. ![](images/join_1.png) We will join with the original data set based on `TailNum` and calculate the plane-adjusted measure of speed. ```r right_join(flights,averages,by="TailNum") %>% mutate(norm_speed = speed / mean_speed) %>% group_by(DayOfWeek) %>% select(contains("speed")) %>% summarise_each(funs(mean(.,na.rm=TRUE))) ``` ``` ## Source: local data frame [7 x 4] ## ## DayOfWeek speed mean_speed norm_speed ## 1 1 419.2984 420.5830 0.9965149 ## 2 2 418.2220 420.4509 0.9942130 ## 3 3 419.2926 420.4236 0.9968347 ## 4 4 420.3597 420.3783 0.9995784 ## 5 5 419.9916 420.4688 0.9985305 ## 6 6 427.3059 422.7704 1.0107539 ## 7 7 423.7673 421.4198 1.0053550 ``` So if things worked out like we expected it looks like Saturday flights are fastest and Tuesday the slowest. ### Using dplyr programmatically Most of the examples up to this point featured using `dplyr` in interactive mode. However there are variants of nearly every verb which are best suited for use inside other functions. To see what theses are take a look at `verb_` versions of each function (e.g. `summarise_`). ### Benchmarking split-apply in base and dplyr Finally I will wrap with a relatively non-sophisticated benchmarking head-to-head comparison of `base` and `dplyr` speed for the almighty split-apply strategy. Lets set up the data. ```r rows<-10000 cols<-100 groups<-100 samples<-rows/groups tmp.data<-data.frame(matrix(rnorm(rows),rows,cols)) tmp.data$group<-rep(1:groups,each=samples) ``` This data set has 10^{4} rows, 100 columns and 100 groups with 100 samples each. `base` ![](images/sad_face.png) ```r ptm <- Sys.time() #split the data on the number of cylinders big.l<-split(tmp.data,tmp.data$group) #apply some function of interest to all columns results<-lapply(big.l, function(x) apply(x,2,median)) #bind results and add splitting info results<-data.frame(group=names(results),do.call("rbind",results)) #elapsed time (bd<-Sys.time()-ptm ) ``` ``` ## Time difference of 0.5802019 secs ``` `dplyr` ![](images/happy_face.gif) ```r ptm <- Sys.time() results<-tmp.data %>% group_by(group) %>% summarise_each(funs(median(.))) #elapsed time (ad<-Sys.time()-ptm ) ``` ``` ## Time difference of 0.3492 secs ``` Wow I just saved 0.2310019 seconds of my life! Create a benchmark visualization comparing base to dplyr for differing number of groups, rows and columns. Uncomment the code in the appendix below and modify as needed to re-run the benchmark. ```r load(file="benchmark results") #create a plot library(reshape2) library(ggplot2) tmp.data<-melt(res,id.vars=c("rows","columns","groups","samples")) %>% mutate(seconds=value*60) ggplot(tmp.data, aes(y=seconds,x=groups,group=variable,color=variable)) + geom_line() +geom_point()+ facet_grid(rows ~ columns) +scale_y_log10() ``` ![](hands_on_with_dplyr_files/figure-html/unnamed-chunk-31-1.png) The plot above shows the calculation time for 10 replications in seconds (y-axis) for calculating the median of varying number of groups (x-axis), rows (y-facet) and columns (x-facet). ***** ### Gotchas 1. In `dplyr` rownames are a second class citizen and are not stored. 2. Even though nearly every `dplyr` tutorial features pipes (`%>%`) it may be easier to learn both `dplyr` and `%>%` separately. 3. Most `dplyr` functions only work on objects coercible to ~ `data.frames`. Lots of my debugging sessions start with trying to understand the data structure of objects I am passing to `dplyr`. ***** ### Appendix ```r # # #set up functions to time # base_fun<-function(data){ # #split the data on the number of cylinders # big.l<-split(data,data$group) # # #apply some function of interest to all columns # results<-lapply(big.l, function(x) apply(x,2,median)) # # #bind results and add splitting info # data.frame(group=names(results),do.call("rbind",results)) # } # # dplyr_fun<-function(data){ # # data %>% group_by(group) %>% summarise_each(funs(median(.))) # } # # #benchmark function # benchmark_fun<-function(rows,cols,groups){ # # #set up data # samples<-floor(rows/groups) # tmp.data<-data.frame(matrix(rnorm(rows),rows,cols)) # tmp.data$group<-rep(1:groups,length.out=rows) # # #base # base.time<-system.time(replicate(10,base_fun(tmp.data))) # # #dplyr # dplyr.time<-system.time(replicate(10,dplyr_fun(tmp.data))) # # data.frame(rows=rows,columns=cols,groups=groups,samples=samples,base=signif(base.time["elapsed"]/10,3),dplyr=signif(dplyr.time["elapsed"]/10,3)) # # } # # #run benchmarks # len<-5 # groups<-seq(5, 100,length.out=len) %>% signif(.,0) # rows<-seq(100, 10000,length.out=len) %>% signif(.,0) # cols<-seq(10, 100,length.out=len) %>% signif(.,0) # # #benchmarks # results<-list() # counter<-1 # for(i in 1:length(groups)){ # .group<-groups[i] # for(j in 1:length(rows)){ # .row<-rows[j] # for(k in 1:length(cols)){ # .col<-cols[k] # results[[counter]]<- benchmark_fun(.row,.col,.group) # counter<-counter+1 # } # } # } # # # res<-do.call("rbind",results) # save(res,file="benchmark results") # #create a plot # library(reshape2) # library(ggplot2) # tmp.data<-melt(res,id.vars=c("rows","columns","groups","samples")) %>% mutate(seconds=value*60) # # ggplot(tmp.data, aes(y=seconds,x=groups,group=variable,color=variable)) + geom_line() +geom_point()+ facet_grid(columns ~ rows) +scale_y_log10() ``` ================================================ FILE: README.html ================================================ README

Teaching Demos

Bioinformatics related demos and tutorials using the R programming language.

R Code Examples and More

Statistical Analysis

Principal Components Analysis (PCA)

PLS

Database query and translation

Network Visualizations

================================================ FILE: README.md ================================================ # Teaching Demos *** Bioinformatics related demos and tutorials using the R programming language. *** # Courses *** ### [see all courses](https://creativedatasolutions.github.io/CDS.courses/) ![](https://creativedatasolutions.github.io/R_programming_for_DS_beginner/imgs/course_logo.png) * ### [R programming for Data Science - a beginner's guide](https://creativedatasolutions.github.io/R_programming_for_DS_beginner/) * ### [Network mapping 101](https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/) *** # Blog ![](imgs/blog.png) * ### [Creative Data Science Blog](https://creativedatasolutions.github.io/cds.blog/) * ### [R programming, bioinformatics and data visualization](https://imdevsoftware.wordpress.com/) *** ## Software ![](imgs/CDS.png) * ### [CDS](https://creative-data.science/#) *** # R Code Examples and More *** ## Data Analysis Examples * ### [Metabolomic and Proteomic Data Analysis 2015](https://github.com/dgrapov/TeachingDemos/blob/master/Demos/Data%20Analysis%20Workflow/report/report.md) *** ## Data Wrangling * ### [Data wrangling for Data Science](https://creativedatasolutions.github.io/R_programming_for_DS_beginner/data_wrangling.html) * ### [dplyr tutorial: verbs and split-apply](https://github.com/dgrapov/TeachingDemos/blob/master/Demos/dplyr/hands_on_with_dplyr.md) *** ## Statistical Analysis * ### [Covariate Adjustment](https://github.com/dgrapov/TeachingDemos/blob/master/Demos/Statistical%20Analysis/Covariate%20Adjustment.md) *** ## Principal Components Analysis (PCA) * ### [PCA for data exploration](https://creativedatasolutions.github.io/R_programming_for_DS_beginner/eda.html#multivariate-analysis) * ### [PCA and scores visualization on custom data](https://gist.github.com/dgrapov/2fd3027d2bd7f15884b2d40a82d50c06) (fast and simple) * ### [PCA on mtcars](https://github.com/dgrapov/TeachingDemos/wiki/Principal-Components-Analysis) (fast and simple) *** ## PLS * ### [Partial Least Squares (PLS)](https://github.com/dgrapov/TeachingDemos/wiki/Partial-Least-Squares) (fast and simple) *** ## Predictive Modeling * ### [O-PLS Example](https://github.com/dgrapov/TeachingDemos/blob/master/Demos/Predictive%20Modeling/O-PLS/OPLS_example.md) (04/08/15) * ### [O-PLS-DA compared to Random Forest](https://github.com/dgrapov/TeachingDemos/blob/master/Demos/Predictive%20Modeling/Iris%20O-PLS-DA/O-PLS_modeling_of_Iris_data.md) (updated 11/7/15) * ### [Comparing single to multiple Y models](https://github.com/dgrapov/TeachingDemos/blob/master/Demos/OPLS/OPLS%20example.md) *** ## Database query and translation * ### [Translating Between Chemical Identifiers](https://github.com/dgrapov/TeachingDemos/wiki/Translating-Between-Chemical-Identifiers) (intermediate) * ### [CTSgetR](https://github.com/dgrapov/CTSgetR) *** ## Network Visualizations * ### [Network mapping 101](https://creativedatasolutions.github.io/CDS.courses/courses/network_mapping_101/docs/) * ### [Building Biochemical and Chemical Similarity Networks](https://github.com/dgrapov/TeachingDemos/wiki/Biochemical-and-Chemical-Similarity-Networks) (advanced) * ### [KEGG Pathway Visualization](https://github.com/dgrapov/TeachingDemos/blob/master/Demos/Pathway%20Analysis/KEGG%20Pathway%20Enrichment.md) (genes, proteins and metabolites) *** ## Miscellaneous * ### [Summer 2014 Metabolomic Data Analysis Workshop](http://imdevsoftware.wordpress.com/2013/09/08/sessions-in-metabolomics-2013/) * ### [Winter 2014 Metabolomic Data Analysis Workshop](http://imdevsoftware.wordpress.com/2014/02/17/tutorials-statistical-and-multivariate-analysis-for-metabolomics/) *** ================================================ FILE: TeachingDemos.Rproj ================================================ Version: 1.0 RestoreWorkspace: Default SaveWorkspace: Default AlwaysSaveHistory: Default EnableCodeIndexing: Yes UseSpacesForTab: Yes NumSpacesForTab: 2 Encoding: UTF-8 RnwWeave: Sweave LaTeX: pdfLaTeX ================================================ FILE: style.R ================================================ options(rstudio.markdownToHTML = function(inputFile, outputFile) { require(markdown) markdownToHTML(inputFile, outputFile, stylesheet=system.file("misc", "docco-template.html", package = "knitr")) } )