I have a list of N genes. I used DAVID (http://david.abcc.ncifcrf.gov) in order to get the GO terms that are statistically overrepresented in that list. I selected the GO_BP_FAT subset of terms to not take in account some really generic GO terms (high level terms in the GO tree). Surprisingly I found out something unclear. Given a specific GO term I have a list of genes (defined by DAVID) that should be statistically associated with that term. Now, if I query the Gene Ontology database using the "official" AMIGO interface (http://amigo1.geneontology.org/) I found that the GO term indicated by DAVID is not associated with some genes. For example I found that PAX3, with other genes, is associated with GO:004582, but if you query Amigo you don't get this term. How is this possible?
As far as I remember, there could be an issue with GO version in DAVID, which was not updated for quite a long time.
From the website it seems that the last update was on 2009 (5 years ago). I don't know exactly how the association between GO terms and gene works, but this means that some associations might have been deleted across the years right?
Yep.. deleted, added, changed. The amigo page actually mentions "GO database release 2014-10-18", so you're right about 5 year difference
if they really did not update for 5 years it would be mind bending
I would recommend using a different tool, like ermineJ
When I started doing bioinformatics a year ago,
davidwasted a lot of my time because of these kind of issues and yet it was mentioned every where as one of the best GEA tools.Gene Ontology Consortiumis definitely a better choice.