GSEA after MACS2 peak calling for ChIP-seq; want positive enrichment only

Question

[sonark]

Hello, I have called histone modification ChIP-seq peaks compared to a control using MACS2 and annotated them using Homer.

I then try to perform pre-ranked GSEA. I only have positive values in my ranking, but I also get negative enrichment so I think GSEA is treating the genes at the bottom of my list as negatively enriched. But they are not; they are just lower but positive. Is there a solution for this? I tried choosing the 'classic' option rather than 'weighted' and it doesn't do it.

Also, what would be the best scoring for pre-ranking? I have fold enrichment, peak score, q-value, p-value and pileup value for each gene.

Answer 1

[alserg]

Development version of fgsea (from github) supports one-tailed tests. However, we don't see much of a difference compared to a standard approach. See here in more details: https://github.com/ctlab/fgsea/issues/27

Answer 2

[sonark]

---

.

Update: I have found that the best & easiest option is to use the "Investigate Gene Sets" page on the GSEA MSigDB website where you can compute overlaps of your gene list with MSigDB gene sets (up to 2026 genes allowed).

It provides a P value based on "Hypergeometric distribution (k, K, n, N) of overlapping genes over all genes in the gene universe."

.

---