[BioC] LIMMA decideTests result zero from contrast matrix

[Alex Gutteridge]

On 09.05.2012 04:21, Ekta Jain wrote:
> Dear Jim,
> I did change things around when you pointed out for the first time.
> All I have been doing is
>> numGenes <- rownames(eset)
>> topTable(fit2, coef=1, adjust="BH", sort.by="B", number=numGenes)
> And
>> results<-decideTests(fit2, method ="global", lfc =0)
> As you mention in your email that "by default decideTests() uses a BH
> adjusted p-value to filter genes" so am i not applying the same
> adjustment for both the toptable() and decideTests() here?
>
> This is what I am clueless about since I still get zero genes. I
> cannot seem to figure out how to not let decideTests use a BH adjust
> for p-value. For the sake of detail, the code worked fine for all my
> cell lines treated with CPI since the p-values were not as bad as the
> ones for treatment with DMSO.
>
> Thank you,
> Ekta

I think it is the difference in the default P value cutoffs for 
topTable and decideTests that is confusing the issue. From the 
decideTests docs:

Usage:

      
decideTests(object,method="separate",adjust.method="BH",p.value=0.05,lfc=0)

Look at the default p.value cutoff - 0.05. Earlier in the thread you 
said your adjusted P values were ~0.9, hence *nothing* will come through 
the filter. The default for topTable is 1:

Usage:

      topTable(fit, coef=NULL, number=10, genelist=fit$genes, 
adjust.method="BH",
               sort.by="B", resort.by=NULL, p.value=1, lfc=0, 
confint=FALSE)

If you want to apply the filter to unadjusted P values the docs say 
this:

adjust.method: character string specifying p-value adjustment method.
           Possible values are ‘"none"’, ‘"BH"’, ‘"fdr"’ (equivalent to
           ‘"BH"’), ‘"BY"’ and ‘"holm"’. See ‘p.adjust’ for details.

Though I'm not sure why you would want to do this.

-- 
Alex Gutteridge