Hi BioC, I am getting weird results from analyzing a set of microarray datasets to find differentially expressed genes. Background: basically I have two strains, mutant and wildtype. And I also have a treatment an inducing factor. So in the end, I have 4 conditions: wildtype, mutant, wildtype induced, mutant induced The way I setup the model is to have 2 factors: STRAIN (wildtype vs mutant) and TREATMENT (induce vs not induce), and I also want to model the interaction between the two. In the end, I want to know which genes show a difference between WT vs MT, induce vs not induce, and respond differently in mutant compared to wild-type cells (i.e. all 3 coefficients are significant). My code essentially is:> contrasts(STRAIN)<-contr.sum(2);> contrasts(TREATMENT)<-contr.sum(2);> design<-model.matrix(~STRAIN*TREATMENT);> fit<-lmFit(allDatagcRMA,design);> cont.matrix<-cbind(WTvsMT=c(0,4,0,0) ,YesvsNoinduce=c(0,0,4,0),Interaction=c(0,0,0,4));> fit2<-contrasts.fit(fit,cont.matrix);> fit2<-eBayes(fit2);> results<-decideTests(fit2); When I viewed my results on vennDiagrams:> vennDiagram(results, include="both");> vennDiagram(results,include=c("up","down"),counts.c ol=c("red","green"))I found the sum of the genes up-regulated and down-regulated don't equal to result from the first diagram. For example, the area that overlaps all 3 circles (i.e. the innerest portion) in the first venndiagram don't equal to the sum of the up- regulated and down-regulated genes in that same area.I am wondering is this because of the way I setup my model? Is my model doing what I intend it to do? Thank you!Sincerely,Casper _________________________________________________________________ [[alternative HTML version deleted]]

Hi Casper, On 7/8/2010 7:26 PM, Casper Shyr wrote: > > Hi BioC, I am getting weird results from analyzing a set of > microarray datasets to find differentially expressed genes. > Background: basically I have two strains, mutant and wildtype. And I > also have a treatment an inducing factor. So in the end, I have 4 > conditions: wildtype, mutant, wildtype induced, mutant induced The > way I setup the model is to have 2 factors: STRAIN (wildtype vs > mutant) and TREATMENT (induce vs not induce), and I also want to > model the interaction between the two. In the end, I want to know > which genes show a difference between WT vs MT, induce vs not induce, > and respond differently in mutant compared to wild-type cells (i.e. > all 3 coefficients are significant). My code essentially is:> > contrasts(STRAIN)<-contr.sum(2);> > contrasts(TREATMENT)<-contr.sum(2);> > design<-model.matrix(~STRAIN*TREATMENT);> > fit<-lmFit(allDatagcRMA,design);> > cont.matrix<-cbind(WTvsMT=c(0,4,0,0),YesvsNoinduce=c(0,0,4,0),Intera ction=c(0,0,0,4));> > fit2<-contrasts.fit(fit,cont.matrix);> fit2<-eBayes(fit2);> > results<-decideTests(fit2); When I viewed my results on > vennDiagrams:> vennDiagram(results, include="both");> > vennDiagram(results,include=c("up","down"),counts.col=c("red","green "))I > found the sum of the genes up-regulated and down-regulated don't > equal to result from the first diagram. For example, the area that > overlaps all 3 circles (i.e. the innerest portion) in the first > venndiagram don't equal to the sum of the up-regulated and > down-regulated genes in that same area.I am wondering is this because > of the way I setup my model? No, it is a misunderstanding of what is meant by 'both' in vennDiagram. You are assuming that 'both' means that both comparisons are up, or both comparisons are down. But what 'both' means here is that both comparisons are significant, without regard to direction. So when you do the Venn diagrams using 'up' and 'down' arguments, you are missing those genes that are up-regulated in one comparison, but down-regulated in the other (which *will* be counted when using the 'both' argument). Best, Jim Is my model doing what I intend it to > do? Thank you!Sincerely,Casper > _________________________________________________________________ > > > [[alternative HTML version deleted]] > > _______________________________________________ Bioconductor mailing > list Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor Search the > archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor -- James W. MacDonald, M.S. Biostatistician Douglas Lab University of Michigan Department of Human Genetics 5912 Buhl 1241 E. Catherine St. Ann Arbor MI 48109-5618 734-615-7826 ********************************************************** Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues