6 months ago by
Spain/Barcelona/Universitat Pompeu Fabra
It's unclear to me to what Bioconductor package are your referring to when you talk about the "genecodis3 tools" as a text-free search among the software packages does not give any hit.
About the general question on how to select enriched GO terms, there are many options, just think for instance that you're talking about the top-10 GO terms without defining what is the metric you use to rank them.
If you're going to select GO terms on the basis of multiple test correction on a one-tailed Fisher's exact test p-value, which is a tricky business given the overlapping nature of GO terms and the consequently lack of independence of the raw p-values assumed by correction methods such as FDR, I would avoid testing GO terms annotated to fewer than 5 and more than 300 genes in your data set. This should alleviate the burden of the multiple test correction, while likely discarding GO terms for which either the representation in your data is too small (< 5 genes) to do a reliable inference, or too large (> 300) to draw useful conclusions. You may of course use different thresholds than the ones I'm proposing.
Regarding the way in which you may rank GO terms, once you've selected them, i'd use the odds-ratio of enrichment.