Questions about GOstats.............

[wetSEQer]

Hello folks.....

I recently ran into a few question regarding the GOstats (Bioconductor package) output.

Here is the parameters for the enrichment analysis:
params<-new('GOHyperGParams',
geneIds=go_object, ###gene of interest
universeGeneIds=universe, ###background genes
ontology='BP',
pvalueCutoff=0.001,
conditional=F,
testDirection='over',
annotation="org.Hs.eg.db")
hgOver<-hyperGTest(params)

When I am running the

probeSetSummary(hgOver) to check which genes are in each ontology category, I ran into this warning message:
Error in (function (classes, fdef, mtable) :

unable to find an inherited method for function ‘columns’ for signature ‘"function"’

If you have a solution, please let me know.

Also, I noticed if a GO_ID is enriched, and the output tells you there are saying 20 counts, I always see more genes with the given go_id from the input list....

Anyone know why?

Thanks

[wetSEQer]

We recently ran into a problem of GOstats package:

Below is how we performed GO enrichment analysis, both the background and input genes are vector of characters containing entrez gene ID.
###set up backgrourd genes
x<-org.Hs.egGO
mapped_genes<-mappedkeys(x)
###set up input genes
go_object<-sample_to_be_test
###Enrichment analysis:
class(go_object)<-"character"
universe<-mapped_genes
params<-new('GOHyperGParams',
geneIds=go_object,
universeGeneIds=universe,
ontology='BP',
pvalueCutoff=0.001,
conditional=F,
testDirection='over',
annotation="org.Hs.eg.db") #for human
hgOver<-hyperGTest(params)
#hgOver #remove # for pipeline checking
result<-summary(hgOver)
After the run, we want to see what genes are in each enriched go_term.
First of all, probeSetSummary(hgOver) does not work, it returned warning message:
Error in (function (classes, fdef, mtable) :
unable to find an inherited method for function ‘columns’ for signature ‘"function"’
Then we converted the input entrez_ID to go_id to manually sort out the genes in each category.

We found the "count" for a given go_id in result doesn't fit with the number of genes carrying the go_id.

Is there a solution to this? If the count is wrong, is the statistic analysis after the gene counting still useful.