How to do pathway analysis after scanpy for single cell data after DE analysis?

Question

I have done DE analysis using SCANpy on my single cell data and I have compared each cluster versus all the other clusters. One cluster seems particularly interesting so I wanted to do pathway analysis.

I have already tried to do the pathway analysis but I got confused. I initially followed the instructions from here (https://nbisweden.github.io/workshop-scRNAseq/labs/compiled/scanpy/scanpy_05_dge.html) and did the enrichment and then I also wanted to do GSEA so I used gseapy.prerank. However, then I was trying to find a way to visualize the network so I ended up going to the gseapy documentation.

The steps of the enrichment analysis seemed the same but then I got confused with the GSEA portion. The first tutorial I found used gseapy.prerank but the second one uses gp.gsea. I was trying to understand the difference and it seems that since I already have the DE analysis I should be using the gseapy.prerank right bit I am not sure.

I was wondering if doing the analysis trough KEGG would be easier.

Thank you

Gene set enrichment analysis (GSEA) is a commonly used algorithm for characterizing gene expression changes. However, the currently available tools used to perform GSEA have a limited ability to analyze large datasets, which is particularly problematic for the analysis of single-cell data. To overcome this limitation, we developed a GSEA package in Python (GSEApy), which could efficiently analyze large single-cell datasets.

Answer 1

An easy way to do an enrichment analysis in scanpy is using scanpy.queries.enrich(). This relies on gprofiler, which uses KEGG, GO and other ontologies

Answer 2

The original question was GSEApy vs KEGG. The response below is GSEApy. The question has been edited and is now a different question concerning options within GSEApy.

---

Response to original question

If GSEApy as described in the publication provides the output you require - that is much, much simpler. KEGG will be complex and you'd need its ids.

However, surely GSEA will require to perform the full calculation of fold-change from scratch rather than be fed the output of scanpy? KEGG will require the output of scanpy. GSEApy primary function is fold-change rather identifying the precisely which pathways are regulating the phenotype under analysis.

KEGG will provide a more accurate answer - in my humble opinion. GSEApy is a lot easier because it very much an "all-in-one" solution. If the answer isn't satisfactory you can always switch.

---

Just to note the GSEApy version I'm referring to is the latest publication.

NOTE There was a deleted post from the OP about an option within GSEA. This is a separate question in the SE format of questions and answers. Thats the way the site works.