correlation analysis
1
0
Entering edit mode
Helen Smith ▴ 100
@helen-smith-6087
Last seen 9.6 years ago
Hi All, I have a general bioconductor question. I've been struggling with how to do this all week, so any help would be much appreciated, I have used correlation analysis in the past to define a range of genes which correlate with a certain prognosis score (all prognosis scores known). Please see the script at the bottom of this page that I have previously applied. This new data set is a bit different as I know the prognosis of 2/3's of the samples but the third set is unknown and I want to figure out individual samples within this third set are more like sample set A or B (different samples within this 'unknown set C' may be split, they are not all necessarily the all the same (good or bad)): * Sample set A o poor prognosis * Sample set B o good prognosis * Sample set C o Unknown prognosis Can the script be amended in some way to account for this and cluster samples within group C near set A or B within the heatmaps? Thanks again, Helen Previous script: setwd("mnlkn") dat<-read.csv("correlation analysis- gene names (SIBLING REMOVAL).csv",header=T) dat[1:10,] Gene<-dat[,1] dat<-dat[,-1] blastocyst<-dat[,1:10] ###################################################################### ############################## expression<-matrix(NA,nrow(sample1),ncol(sample1)) ####Take Average logfc of multiples of the same gene#### gid<-unique(Gene) mGene<-matrix(NA,length(gid),ncol(sample1)) for(i in 1:length(gid)){ rid<-which(Gene==gid[i]) for(j in 1:ncol(blastocyst)){ mGene[i,j]<-mean(blastocyst[rid,j]) } } mGene[1:10,1:10] dim(mGene) t.mGene<-t(mGene) t.mGene[1:10,1:10] ###################################################################### ############################## ######prog scores and ranks for All 15 samples, with patient duplicates) ###pearson### age<-c(33,28,37,41,38,37,33,33,28,36) ###spearman### age.rank<-c(2,1,4,6,5,4,2,2,1,3) ###################################################################### ############################## COR_3<-rep(NA,19992) COR_4<-rep(NA,19992) for(i in 1:19992){ COR_3[i]<-cor(age.rank,rank(abs(t.mGene[,i])),method="spearman") COR_4[i]<-cor(age,(abs(t.mGene[,i]))) } par(mfrow=c(2,1)) plot(COR_3,ylim=c(-1,1));abline(h=c(0.75,-0.75),lwd=2,col=2,lty=2) plot(COR_4,ylim=c(-1,1));abline(h=c(0.75,-0.75),lwd=2,col=2,lty=2) ###################################################################### ### ###################################################################### ### ##########################SPEARMANS RANK################################# ####Setting a series of different thresholds THRESH<-seq(0.7,0.8,by=0.01) par(ask=TRUE) THRESH #ranked prognosis scores, aka COR_3#### for(k in 1:10){ thresh<-THRESH[k] rank.abs.2<-which(abs(COR_3)>thresh) rank.pos.2<-which(COR_3>thresh) rank.neg.2<-which(COR_3<(-thresh)) ### Creating labels for plot ## Gene[rank.neg.2] Prog.id<-c("good","bad","good","good","good","bad","good","bad","bad", "bad") ###### id<-rank.neg.2 ### heat map ht.dat<-as.data.frame(mGene[id,]) names(ht.dat)<-Prog.id row.names(ht.dat)<-gid[id] ht.dat<-as.matrix(ht.dat) heatmap(ht.dat)} ###################################################################### #### ##plot specfic thresholds for powerpoint before and after breaks down ### Setting threshold for boundary of interest ############absolute thresh<-0.5 rank.abs.2<-which(abs(COR_3)>thresh) id<-rank.abs.2 ### heat map ht.dat<-as.data.frame(mGene[id,]) names(ht.dat)<-Prog.id row.names(ht.dat)<-gid[id] ht.dat<-as.matrix(ht.dat) heatmap(ht.dat) ht.dat ####identify genes with greater expression than 50 (log 5.6438) new.mat<-ht.dat new.mat[(which(abs(new.mat)<5.6438))]<-"" as.numeric(new.mat) t(new.mat) ## Getting the Genes of interest GENES<-gid[id];GENES ## boxplot par(mfrow=c(1,1)) par(las=2) boxplot(ht.dat) genes<-gid[id]###lists names of genes on heatmap genes length(genes) ##########positive thresh<-0.76 rank.pos.2<-which(COR_3>thresh) id<-rank.pos.2 ### heat map ht.dat<-as.data.frame(mGene[id,]) names(ht.dat)<-Prog.id row.names(ht.dat)<-gid[id] ht.dat<-as.matrix(ht.dat) heatmap(ht.dat) ht.dat ####identify genes with less greater than 2fc (lof1) new.mat<-ht.dat new.mat[(which(abs(new.mat)<1))]<-"" as.numeric(new.mat) t(new.mat) ## Getting the Genes of interest GENES<-gid[id];GENES [[alternative HTML version deleted]]
• 770 views
ADD COMMENT
0
Entering edit mode
@steve-lianoglou-2771
Last seen 14 months ago
United States
Hi, On Fri, Feb 28, 2014 at 4:18 AM, Helen Smith <helen.smith-2 at="" manchester.ac.uk=""> wrote: > Hi All, > > > > I have a general bioconductor question. I've been struggling with how to do this all week, so any help would be much appreciated, > > > > I have used correlation analysis in the past to define a range of genes which correlate with a certain prognosis score (all prognosis scores known). Please see the script at the bottom of this page that I have previously applied. > > > > This new data set is a bit different as I know the prognosis of 2/3's of the samples but the third set is unknown and I want to figure out individual samples within this third set are more like sample set A or B (different samples within this 'unknown set C' may be split, they are not all necessarily the all the same (good or bad)): > > * Sample set A > > o poor prognosis > > * Sample set B > > o good prognosis > > * Sample set C > > o Unknown prognosis > > > > Can the script be amended in some way to account for this and cluster samples within group C near set A or B within the heatmaps? This is a "classic" statistical/machine learning problem. You have a set of samples with known labels, and you want to build a machine that learns how the features of each sample relate to its label. You'd then want to take this "machine" and apply it to your new data. All of the details regarding how this is done is not appropriate to communicate over email, but you're in luck. Trevor Hastie and Rob Tibshirani are currently teaching a MOOC that covers the details of this stuff at an introductory level. Even better, they make their videos available for download, so you really ingest the details at your own pace, and even better (++) all of their stuff is in R: https://class.stanford.edu/courses/HumanitiesScience/StatLearning/Wint er2014/about -steve -- Steve Lianoglou Computational Biologist Genentech
ADD COMMENT

Login before adding your answer.

Traffic: 849 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6