[BioC] LIMMA, FDR and B-statistic

[Jordi Altirriba Gutiérrez]

Hello to everyone!!
I've been using RMA to normalize my data and LIMMA to obtain a list of 
significant genes. My design was a 2x2 factorial design with 4 groups: 
Diabetic treated, diabetic untreated, health treated and health untreated 
with 3 biological replicates in each group.
I've got the list of significant genes with these commands (thanks again to 
Gordon!):
>design<-model.matrix(~DIABETES*TREATMENT,data=pData(eset))
>fit<-lmFit(eset,design)
>contrast.matrix<-makeContrasts(DIABETESTRUE,TREATMENTTRUE,DIABETESTRUE.TREATMENTTRUE,levels=design)
>fit2<-contrasts.fit(fit,contrast.matrix)
>fit2<-eBayes(fit2)
>topTable(fit2, 
>number=100,genelist=geneNames(eset),coef="DIABETESTRUE",adjust="fdr")
Now I've more questions (sorry to bother you all again).
1.- Is it possible to know at what false discovery rate are we working with 
these 100 genes? (something similar to the median and the 90th percentile of 
FDR that we obtain with SAM). If so, how can I get to know it ?
2.- When I observe my genelist for the TREATMENT I realize that the first 
gene of the list has a negative B value (-2.83), however when I obtain the 
genelist for the TREATMENT.DIABETES, in this case what I get for the top 
gene is a B value of 14. Is it correct to interpret that the drug only acts 
in the diabetic animals and in the healthy ones does not induce any 
difference in the gene expression?
3.- When we work with the p-value, there is an agreement (more or less) that 
a value <0.05 is significant. Is there an agreement with the B statistic? 
(I've read "replicated microarray data" of Lönnstdt and Speed and I think 
that it depends on your data and experiment, but is there any way to 
determine the cutoff?)
Thanks again for your suggestions and patience!
Yours sincerely,

Jordi Altirriba, PhD student
IDIBAPS - Hospital Clinic (Barcelona, Spain)

_________________________________________________________________
Get tax tips, tools and access to IRS forms – all in one place at MSN Money!