Topgo Vs Gostats

Question

[enricoferrero]

Hello Biostar,

I'm performing some Gene Ontology enrichment analysis using R/Bioconductor and, as far as I understand, there are two packages that allow to do this: topGO and GOstats (I'm not interested in packages that compare multiple datasets in this scenario).

The main difference seems to be the statistics used: GOstats uses the hypergeometric distribution (basically the standard way to test for overrepresentation), while topGO allows the user to use a much wider and complete range of algorithms and statistical tests to check for enrichment.

With regards to graphic capabilities, topGO seems to be producing much better graphs than GOstats, but I have only looked at the packages' vignettes.

Has anybody here experience with both? Which one did you end up using? Why?

Also, on a related note, what would be the topGO algorithm/statistics test combination that emulates GOstats' hypergeometrical test? Would it be the classic/fisher combination (used as the most basic example in topGO's vignette)?

Thank you.

Answer 1

[nitsuaq]

To avoid the following situation:

I tried both GOstats and topGO briefly before switching to topGO for all of my GO enrichment analysis in R. Its been a while but from what I remember it seemed more versatile, i.e. one can employ various default algorithm/test combinations (along with the capability of customized tests).

It seems reasonable that the classic/fisher would emulate GOstats hypergeometric test, though I have not tested this personally.

I found this was a good resource for working with topGO.