[BioC] hyperGTest(GOstats) result question

[Seth Falcon]

burak kutlu <burak_kutlu at yahoo.com> writes:
>>But the GO::GOALLENTREZID map is odd in that it contains data across
>>different organisms.  Why not instead:
>
>> affyWithMyGO = hgu133plus2GO2ALLPROBES[["GO:0009790"]]
>> egWithMyGO = unique(unlist(mget(affyWithMyGO, hgu133plus2ENTREZID)))
>
> Intersection of 'egWithMyGO' with 'universeGenes' contains 31 genes.
>
>> mygenes = intersect(egWithMyGO, universeGeneIds) 
>
> When I run the hyperGTest...
>
>> params = new("GOHyperGParams", geneIds = mygenes, universeGeneIds = universeGeneIds,
>              annotation = "hgu133plus2",ontology = "BP", pvalueCutoff = 0.05,conditional = FALSE,testDirection = "over")
>
>> hyp = hyperGTest(params)
>> t = summary(hyp)
>
> I get:
>
>> t["GO:0009790",]
>                GOBPID       Pvalue OddsRatio  ExpCount Count Size
> GO:0009790 GO:0009790 4.455886e-60       Inf 0.9285024    31   31
>                             Term
> GO:0009790 embryonic development
>
>
> This result does make sense, note that the Count and Size are the
> same, 31. We used 31 genes as input, got 31 back. So my question
> about hyperGTest is answered, I was just confused with the different
> number of genes.
>
> Then my new question is why does 'GOALLENTREZID' contain only 8
> genes associated with GO:0009790 while 'egWithMyGO' obtained from
> 'hgu133plus2GO2ALLPROBES' contains more, i.e. 31 genes.  One would
> expect at least the same amount of genes in both environments or
> maybe more in 'GOALLENTREZID' because affy chip does not represent
> all the genes?

I'm confused.  Here's what I see with egWithMyGO computed as above:

> egFromGO = unique(GOALLENTREZID[["GO:0009790"]])
> length(egFromGO)
[1] 3963
> all(egWithMyGO %in% egFromGO)
[1] TRUE

I'm using devel versions here, so there could be differences in the
counts, but I really expect the egWithMyGO to be a subset of egFromGO.

-- 
Seth Falcon | Computational Biology | Fred Hutchinson Cancer Research Center
http://bioconductor.org