[BioC] FDR control in conditional Hypergeometric test

[Tarca, Adi]

Hi all,

I have a question about GOstats.
When I do an non-conditional analysis (conditional = FALSE) something like:

 params <- new("GOHyperGParams", geneIds = myDEgenes,  universeGeneIds = myUniverse, annotation = mychip,  ontology = "BP", pvalueCutoff = 1.0, conditional = FALSE,  testDirection = "over")  
hgCondOver <- hyperGTest(params)
tmp<-summary(hgCondOver)

and I want to control the FDR, I use:
 tmp$pFDR<-p.adjust(tmp$Pvalue,"fdr")

since tmp contains all GO terms tested, and therefore I know exactly how many tests were done.

When I use: 
params <- new("GOHyperGParams", geneIds = myDEgenes,  universeGeneIds = myUniverse, annotation = mychip,  ontology = "BP", pvalueCutoff = 1.0, conditional = TRUE,  testDirection = "over")  
hgCondOver <- hyperGTest(params)
tmp<-summary(hgCondOver)

the conditioning will be done always because pvalueCutoff = 1.0, and I can still control the FDR as above. 

And finally the question. How can I do conditioning only in some of the case (e.g. pvalueCutoff = 0.05) but still be able to control the FDR or FWER for the tests that I am doing. Because if I use pvalueCutoff = 0.05, the tmp dataframe will contain only the terms with Pvalue less than 0.05.


Thank you,

Adi