[BioC] hyperGTest, different Results using different annotation packages

[Wiebke Iffert]

Dear All,

I want to do an analysis using the function hyperGTest from pakage GOstats.
When I use the geneIDs of my own annotation package a get other result from the analysis than when using the annotation package org.Hs.eg.db.

I have built my own annotation package using AnnotationDbi following the instructions in the vignette (I'm using data from a self spotted microarray with oligonucleotides which I mapped to Entrez geneIDs):

>library(AnnotationDbi)
>makeHUMANCHIP_DB(affy=FALSE, prefix="HighDensityArray", fileName="high_density_array_oid2eg.txt", >baseMapType="eg", outputDir = getwd(), version="1.0.2", manufacturer = "selfspotted", 
>chipName = "High Density Array", manufacturerUrl = "NA") 

To build an object of class GOHyperGParams, I used the geneIDs corresponding to the oligonucleotides of interest.

>paramsBPover<-new("GOHyperGParams",geneIds= genesOfInterest,universeGeneIds= allgenes, >annotation="HighDensityArray.db", ontology="BP", pvalueCutoff=0.05, conditional=FALSE, >testDirection="over", categoryName="GO") 
>hgOver.BP<-hyperGTest(paramsBPover)

and I got the result:
>hgOver.BP
Gene to GO BP  test for over-representation 
1449 GO BP ids tested (180 have p < 0.05)
Selected gene set size: 203 
    Gene universe size: 1768 
    Annotation package: HighDensityArray.db 


By coincidence I started the same methods using the annotation package org.Hs.eg.db with the same geneIDs:

>paramsBPover2<-new("GOHyperGParams",geneIds= genesOfInterest,universeGeneIds= allgenes, >annotation="org.Hs.eg.db", ontology="BP", pvalueCutoff=0.05, conditional=FALSE, testDirection="over", >categoryName="GO") 
>hgOver.BP2<-hyperGTest(paramsBPover2)

and I got the result:
>hgOver.BP2
Gene to GO BP  test for over-representation 
967 GO BP ids tested (118 have p < 0.05)
Selected gene set size: 203 
    Gene universe size: 1768 
    Annotation package: org.Hs.eg.db 

Shouldn't I get the same results independent from these 2 annotation packages? (I thougth that my package HighDensityArray is something like an subset of the org.Hs.eg.db package, but using my oligonucleotide IDs as Identifier instead of geneIDs - or did I get that wrong?).
Which analysis is the one to rely on?


Thanks in advance for any help.
wiebke


P.S. 
sessionInfo()
R version 2.8.0 (2008-10-20) 
i386-apple-darwin8.11.1 

locale:
de_DE.UTF-8/de_DE.UTF-8/C/C/de_DE.UTF-8/de_DE.UTF-8

attached base packages:
[1] splines   tools     stats     graphics  grDevices utils    
[7] datasets  methods   base     

other attached packages:
 [1] org.Hs.eg.db_2.2.6        GOstats_2.8.0            
 [3] RBGL_1.18.0               GO.db_2.2.5              
 [5] HighDensityArray.db_1.0.2 RSQLite_0.7-1            
 [7] DBI_0.2-4                 Category_2.8.1           
 [9] genefilter_1.22.0         survival_2.34-1          
[11] annotate_1.20.1           xtable_1.5-4             
[13] AnnotationDbi_1.4.3       graph_1.20.0             
[15] Biobase_2.2.1            

loaded via a namespace (and not attached):
[1] GSEABase_1.4.0  XML_1.98-1      cluster_1.11.11


--