In the topGO, when utilizing the elim method with KS, it is unclear whether adjusted p-values are required.

According to the method, (https://academic.oup.com/bioinformatics/article/22/13/1600/193669) it appears as though a correction for multiple comparisons is computed inside the algorithm (using Bonferonni comparison), meaning that a correction outside of the algorithm is not required (barring some small correction because the set of tested genes varies across the ontology terms).

The topGO manual suggests that adjusting for multiple corrections may produce poor results, yet the authors of the above paper utilize a correction when producing their results.

I also realize that GO is primarily for exploratory analysis, so the p-values matter less than the rankings, but it would be nice for reporting purposes.