[BioC] Re: limma package

[Gordon K Smyth]

> Hi,
> I'm the PhD student who asked you about the multiple slide analysis and
> to whom you quickly answer. I wanted to thank you but I still have
> question about your package.
> This time is concerning the toptable() function.
>
> #My data are:
> 3 condition(3 diffrent time) and 3 or 2 slides per condition i.e.
>
> GeneName ID M5.1 M5.2 M5.3 M15.1 M15.2 M30.1 M30.2
>
> "factor"	0	0	0	1	1	2	2
>
> #My code is:
>
> dataM<-read.table("M_total.txt",header=TRUE,sep="\t")
> RatiodataM<-cbind(data[,3:5],data[,6:7],data[,8:8])
> Mtimes<-factor(c(0,0,0,1,1,2,2))
> Mdesign<-model.matrix(~Mtimes-1)
> Mfit<-lm.series(dataM,Mdesign)
> ebM<-ebayes(Mfit)

You aren't estimating anything of interest - you're just estimating the
three group means.  You need to use the function contrasts.fit to define
the comparisons between the times that you're interested in.  I am
guessing that you have Affymetrix microarray data.  There is an example in
the LIMMA User's Guide (in the development version, not the BioC 1.2
release) in the section on Affymetrix data which fits your data almost
exactly.

Gordon

> totalresM<-cbind(dataM,"p.value_differentiel"=ebM$p.value)
> write.table(totalresM,"Limma_analysis_M.txt",sep="\t",row.names=FALSE)
>
> #I think I understand that part but next
>
> toptableM<-toptable(number=nrow(dataM),genelist=dataM[,1:2],fit=fit,eb=eb,ad
> just="fdr")
>
> When I display the toptableM result I get p.adjust value for the first
> class of the design, i.e. c(0,0,0) and the M(mean) display is also the
> mean of the first three M, i.e. the mean of Mtimes.0
> (mean(M5.1,M5.2,M5.3)
> Is this normal ?
> When we do such analysis do we only compare the data where class
> Mtimes(0,0,0) versus Mtimes(1,1,1,2,2,2)? Can we have adjust p.value for
> the Mtimes(1,1,1) and an estimation of the top genes at that time ?
>
> Nolwenn Le Meur
>
>
> ********************************************
> Nolwenn Le Meur
> INSERM U533
> Faculté de médecine
> 1, rue Gaston Veil
> 44035 Nantes Cedex 1
> France
>
> Tel: (+33)-2-40-41-29-86 (office)
>      (+33)-2-40-41-28-44 (secretary)
> Fax: (+33)-2-40-41-29-50
> mail: nolwenn.lemeur at nantes.inserm.fr
> ********************************************