This does not matter, since:

(i) the adjustment is just a monotone transformation and

(ii) you should not test hypotheses, because the enrichment analysis is an exploratory method. The p-value (adjusted or unadjuted) can be used to rank the enrichment scores by increasing "statistical signal-to-noise ratio".

There was need to update p-value to adjusted p-value. In view of that I would recommend the use of the adjusted p-value in preference to the ordinary p-value.

Jochen Wilhelm Thank you for the reply. Please correct me if I am wrong, so you mean that I can choose either of them.

I have a gene list and using Enrichr software, I am trying to look for KEGG pathways and GO. Now after putting my list into the software, KEGG output:100+ pathways, GO_cellular;100 terms, GO_molecular;100+ terms and GO_biological: 100+ terms. Enrichr can rank pathways and terms according to p-value, adjusted p-value, Odds ratio, and combined score. Apparently, I have to rank them somehow to get the top GO terms and Kegg pathways. So what will be the best approach to rank them?

Ette Etuk Thank you for the reply. the only problem with the adjusted p-value, I am getting a few hits only.

The rank order by p-value should be the same as the rank order by adjusted p-value. Monotone transformations don't change the rank oder.

You should really only use these p-values (adjusted or unadjsted) to rank the candidates, so that you have the "best" candidates (according to statistical signal-to-noise ratio) on the top of the list.

You should not use p-values (adjusted or unadjsted) to assess statistical significance and to formally reject null-hypotheses. Although this is frequently done, this is not a sensible approach. Some authors advocat selecting candidates by FDR (what is based on Benjamini-Hochberg adjusted p-values), but the lists presented have no advantage over presenting the list of the "top-n" candidates (ranked by some useful criterion, what may be the p-value [and here is is irrelevant if the p-values are adjusted]).

You usually should not need any "selection" of a sub-list by a statistical criterion. Showing the top-10 or top-20 in a publication iis ok for the reader to get an impression, and the whole list should be published and accessible (in the supplement or some online repository) anyway. To gor further in your research you may want to pick some individual promising candidates. This selection may be guided be p-values (again: it does not matter if you use adjusted or unadjusted p-values), but it should also be guided by some biological understanding and the research context. The enrichment analysis then helped you to generate (!) hypotheses that you may test in subsequent, well-designed experiments (knock-in, knock-out etc).

If your research question is more "holistic", then you should always use all the data and avoid creating "candidate lists" where you have to make a dichotomous decision whether or not each item (pathway, gene, whatever) is selected (perturbed, regulated, ...).

Jochen Wilhelm Thank you so much for this reply. I really appreciate and thankful that you answered the question in such coherent detail. And now I have a better understanding of this whole concept.