How to decide which are the most pertinent overly-represented features during gene set enrichment analysis