Help on matrix design in Limma - paired samples and two conditions
1
0
Entering edit mode
@lujian0311gmailcom-3857
Last seen 9.6 years ago
I am new to Limma and have some basic questions to ask about the matrix design: (1) I have used RNA-Seq methods to get gene expression profiles from 6 patients, three of them are male, and the other three are females. (2) For each patient, we got expresson data of the normal tissues and cancer tissues. I want to use Limma to analyze these RNA-Seq data and want to address the following questions: (1) Differentially expressed genes in Normal and Cancer tissues in Male. (2) Differentially expressed genes in Normal and Cancer tissues in Female. (3) Whether male and females have the same set of genes that are differentially expressed genes between normal and cancer tissues. Thus I have the target object which is as the following: > target gender treatment biolrep 1 F normal 1 2 F cancer 1 3 F normal 2 4 F cancer 2 5 F normal 3 6 F cancer 3 7 M normal 4 8 M cancer 4 9 M normal 5 10 M cancer 5 11 M normal 6 12 M cancer 6 Below is the code for the limma process. I am not sure whether I should treat the paired samples into a block or not. Any suggestions will be greatly appreciated. TS<-factor(TS,levels=c("F.normal", "F.cancer", "M.normal", "M.cancer")) design<-model.matrix(~0+TS) colnames(design)<-levels(TS) corfit <- duplicateCorrelation(eset, design, ndups=1, block=target$biolrep) fit <- lmFit(eset, design, ndups=1, block=target$biolrep, cor=corfit$consensus) contrasts <- makeContrasts(Fdif=F.normal-F.cancer,Mdif=M.normal-M.cancer,Diff=F.no rmal-F.cancerF)-(M.normal-M.cancer),levels=design) fit2 <- contrasts.fit(fit, contrasts) fit2 <- eBayes(fit2) results<-decideTests(fit2,lfc=log2(1.1)) vennDiagram(results) [[alternative HTML version deleted]]
Cancer limma PROcess Cancer limma PROcess • 2.1k views
ADD COMMENT
0
Entering edit mode
@sunny-srivastava-3793
Last seen 9.6 years ago
Hello, I am sure that more experienced members on this list will have better things to say about your design matrix, but here is my $0.25. As a block receives all the treatments in the design (in your case, Normal and Cancer) so each biorep is a block. Further if I understand your problem correctly then- TS<-factor(TS,levels=c("F. normal", "F.cancer", "M.normal", "M.cancer")) design<-model.matrix(~0+TS) makes sense only when you can consider Males and Females to a treatment. Is this justifiable? A more traditional way is to block on sex, but since you already have blocked on a patient, introducing another block is a good idea if you designed your experiment taking this into account (something like latin squares - but the data that you have shown doesn't point to this design) As you are interested in differential expression in males and females separately (which I think is a good way to analyze this data). If I were to construct a design matrix, I would have constructed it separately for males and females. ## for females TS.f <-factor(TS,levels=c("F. normal", "F.cancer")) design.f <-model.matrix(~0+TS.f) ## for males TS.m<-factor(TS,levels=c("M.normal", "M.cancer")) design.m<-model.matrix(~0+TS.m) I know doing this would reduce the sample size and hence related power issues, you can avoid this if the assumption of treating M and F as treatments is justified. Hope this helps! I may be very wrong, but I would be very happy if someone corrected me. Thanks, Sanvesh. On Thu, Dec 17, 2009 at 4:48 PM, <lujian0311@gmail.com> wrote: > I am new to Limma and have some basic questions to ask about the matrix > design: > > (1) I have used RNA-Seq methods to get gene expression profiles from 6 > patients, three of them are male, and the other three are females. > (2) For each patient, we got expresson data of the normal tissues and > cancer tissues. > > I want to use Limma to analyze these RNA-Seq data and want to address the > following questions: > > (1) Differentially expressed genes in Normal and Cancer tissues in Male. > (2) Differentially expressed genes in Normal and Cancer tissues in Female. > (3) Whether male and females have the same set of genes that are > differentially expressed genes between normal and cancer tissues. > > Thus I have the target object which is as the following: > > target > gender treatment biolrep > 1 F normal 1 > 2 F cancer 1 > 3 F normal 2 > 4 F cancer 2 > 5 F normal 3 > 6 F cancer 3 > 7 M normal 4 > 8 M cancer 4 > 9 M normal 5 > 10 M cancer 5 > 11 M normal 6 > 12 M cancer 6 > > Below is the code for the limma process. I am not sure whether I should > treat the paired samples into a block or not. Any suggestions will be > greatly appreciated. > > > TS<-factor(TS,levels=c("F.normal", "F.cancer", "M.normal", "M.cancer")) > design<-model.matrix(~0+TS) > colnames(design)<-levels(TS) > > corfit <- duplicateCorrelation(eset, design, ndups=1, block=target$biolrep) > fit <- lmFit(eset, design, ndups=1, block=target$biolrep, > cor=corfit$consensus) > > contrasts <- > > makeContrasts(Fdif=F.normal-F.cancer,Mdif=M.normal-M.cancer,Diff=(F. normal-F.cancerF)-(M.normal-M.cancer),levels=design) > > > fit2 <- contrasts.fit(fit, contrasts) > fit2 <- eBayes(fit2) > > results<-decideTests(fit2,lfc=log2(1.1)) > vennDiagram(results) > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > [[alternative HTML version deleted]]
ADD COMMENT

Login before adding your answer.

Traffic: 761 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