[BioC] visualizing limma results

[daniela marconi]

Hi to everybody
I have  used limma to find out genes with different expression with high level of confidence in a Multiclass comaparison.

Now I need to visualize by a heatmap our results.
Only for visualization pourposes using functions heatmap and heatmap.2 I 've tried to separete  different classes with the most significant genes. But I didn't succeded!
Is it strange? No? Or may be am I misinterpreting something?

I expected a good separation yet with genes that have a significant adjusted p.value (less then e-5)...but nothing happened...So I tried to filter also by log odds (B>3) and Coefficient (Coeff<-2 and Coeff>2)...but I didn't manage to separate different classes comparing for example sM vs UM.

Have you any suggestions for visualizing limma results with heatmap?Is it incorrect to expect a separation of classes by heatmap with limma selected genes?

Thanks in advance 
Daniela

Here you can find part of my R code:

>RG<-read.maimages(Targets$file.name,source="genepix",wt.fun=wtflags(0.01))
>RG$printer<-getLayout(RG$genes)
>MA<-normalizeWithinArrays(RG,bc.method="minimum")
>MAlast<-normalizeBetweenArrays(MA,method="quantile")
group<-factor(c("nsM",rep("UM",7),"nsM","sM","nsM","sM","nsM","NC",rep("sM",2),"UM","sM","UM","nsM","UM","nsM",
rep("sM",2),"UM","nsM",rep("UM",3),"sM",rep("nsM",2),"sM",rep("nsM",2),"sM","nsM",rep("sM",3),"nsM","sM","nsM",
"UM","sM","UM",rep("sM",5),rep("UM",2),"nsM","UM",rep("nsM",3)),levels=c("nsM","UM","sM","NC"))
>design<-model.matrix(~0+group)
>colnames(design)<-c("nsM","UM","sM","NC")
>cont.matrix<-makeContrasts(UM.M=UM-(nsM+sM),UM.nsM=UM-nsM,UM.sM=UM-sM,nsM.sM=nsM-sM,levels=design)
>dupcor <- duplicateCorrelation(MAlast,design=design)
>fit2COR<-contrasts.fit(fitCOR,cont.matrix)
>fit2COR<-eBayes(fit2COR)
>results<-decideTests(fit2,method="nestedF",adjust.method="BY",p.value=0.001)

#####tab is a summary genereted in the workspace matrix with a function like write.fit , containing all of information of fit2COR
 
>a<-which(tab$Genes.ID!="empty" & tab$Res.UM.sM!=0 & tab$B.UM.sM>2 & tab$p.value.UM.sM<0.000001

>q<-which(MAlast$targets$Cy5=="sM" )
>qq<-which(MAlast$targets$Cy5=="UM")
>index<-c(qq,q)

>heatMATRIX<-as.matrix(MAlast$M[a,index])
>colnames(heatMATRIX)<-paste(MAlast$targets$Cy5[index],MAlast$targets$file.name[index])
>heatmap.2(heatMATRIX,col=redgreen(75), scale="row", key=TRUE, symkey=FALSE, density.info="none", trace="none", cexRow=0.5)