How to interpreting that ORA and GSEA results?

Question

[ubuntubak]

I'm writing a college paper on comparing ORA and GSEA in pathway enrichment analysis. I'm using some bos taurus data as an input dataframe. Also I'm working on KEGG 2019 (ENSBTAG translated to common gene_names).

And I got result where I got more than 200 paths found in GSEA and 81 in ORA. However, when I filtered the GSEA results FDR <= 0.25, and the ORA adjustment p-value <= 0.05 I was left with 21 GSEA results and 1 ORA result.

I was interested in this one ORA pathway and decided to compare it with the result in GSEA.

It turned out that GSEA also found such a pathway, but it has a negative NES and FDR of 0.5.

Could someone explain to me in simple language how to interpret such a result?

From my point of view - albeit I'm completely new to Pathway Enrichment Analysis - it means that GSEA found this pathway, with a high score for NES, so it has reduced expression, but is uncertain, and at the same time ORA confirms to me the significance of this pathway, with a low adjustment p-value score. Is this the correct interpretation?

If so, using only GSEA, would I wrongly omit this pathway from further study? Am I understanding this completely wrong?

This is my row from result (from Term to Genes is ORA, and from Name to Gene % is GSEA):

better resolution on my google drive: https://drive.google.com/file/d/1bklXAerGgbY3okFof-JJ3n1TpjyNc8mk/view?usp=sharing

Answer 1

[Istvan Albert]

You can only compare the methods if you know the ground truth.

You can't just use as you say some bos taurus data for that purpose.

You should approach this problem as a reproducibility issue.

Select genes that are known to form a pathway and see if you can reproduce the known results with both methods.

[ubuntubak]

Istvan Albert thank you for an answer. So if i understand correctly, I need to find ground truth data, where all is clear and known, and it will be something as reference?

And then, based on this data I need to run GSEA and ORA again, pick relevant paths, and check result based on that?

Can you recommend some good source to learn GSEA, ORA and a pathway analysis generally?

[Istvan Albert]

There is no straigthforward answer here, see for example:

• Interpretation of biological experiments changes with evolution of the Gene Ontology and its annotations

Here is another recent tool that for some reason I feel it is quite trustworthy, more of a hunch at this point since I have not used it yet

• https://github.com/ftwkoopmans/goat