Pre-ranked GSEA on a small subset of genes

16 views
Skip to first unread message

Shmuel

unread,
Jun 16, 2024, 8:45:28 AMJun 16
to gsea-help
Hello,
I am working with single-cell data where the capture is low. 
After filtering the genes for minimal expression, I am left with 2000 genes after filtering.
I run a regression on the data, with the value of each gene being the explained variable, the quality I am interested in as the explanatory variable, with one confounding variable.
The  sign( of the coefficient) * -log10(of the pvalue) for the explanatory variable I am interested in is "fed" into pre-ranked GSEA.

My question is whether the small number of genes that my statistical test starts with can cause the FDR of GSEA's results to be artifically smaller, causing false positives?

If by starting with a small subset of genes can only cause the FDR to be artificially larger (causing false negatives) this is not a problem - I am worried that the results that I do see are false positives.

Anthony Castanza

unread,
Jun 20, 2024, 5:14:25 PMJun 20
to gsea-help
Hi Samuel,

Sorry for the delay getting back to you here, I reached out to a colleague of mine who has more experience in this area and this is what they had to say

I think this is much more likely to produce false negatives rather than false positives. It's going to have a systemic bias towards removing genes in the middle of the ranked list (of linear model coefficients) so it will "spread" the gene set genes along the ranked list and deflate the enrichment (and increase the FDRs). I played around a little with simulated linear model coefficients and gene sets with GSEAPreranked and this seemed to be the case. 

Hope this helps. Let me know if you have any other questions

-Anthony

Anthony S. Castanza, PhD
Curator, Molecular Signatures Database
Mesirov Lab, Department of Medicine
University of California, San Diego

--
You received this message because you are subscribed to the Google Groups "gsea-help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gsea-help+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gsea-help/7624b076-19a1-4132-a9a2-6590cd72d930n%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages