Input all the ranked genes into pre-ranked GSEA?

Question

[blairzxy]

Hi~

I would love to run pre-ranked GSEA on a list of genes. May I ask if I should input all of my genes into pre-ranked GSEA? or I should filter it based on some cutoff and input those selected? I'm leaning towards the first option but not so sure.

Thanks ahead if you could help me!!!!

[venu]

Did you check this post and this?

It would be nice if you have a good number of significant genes after multiple correction and fold change cut-offs. If not you can also use all genes to get an idea on how most of your ranked genes are enriched (i.e. enrichment for certain related biological processes or cell-type etc).

[blairzxy]

Wow, the links you've posted are VERY helpful! Thank you very much!!

For others who would love to see the possible answer, here is the content that cited from the second link which venu posted above: "The key here is that you need to supply ALL genes that are detected in the experiment (all probes for microarray and all genes detected in the RNA-seq experiment). " -- I didn't know this before. :)

I guess the answer would be yes, you could input ALL genes into GSEA if you want and just make sure the way you set your rank score is correct. One possible way to set the score is "rank_score=sign_of_FC * -log(pval)". Feel free to comment!

Answer 1

[alserg]

While you indeed can input all your genes into GSEA, personally usually select top 12000 expressed genes. This allows to reduce noise from lowly-expressed and non-expressed genes.