Trying to understand when GSEA is more appropriate than GSVA and vice versa. I have seen cases when running GSEA and GSVA on the same task - compare enrichment of a geneset between two groups - gives different results. In this case one method will identify significant enrichment while the other will not. GSEA seems to detect more "hits" than GSVA.

I understand that they are different methods operating on different levels (group/condition level vs single sample level) and GSVA is needed when you want to investigate the relationship between a geneset and a non-binary grouping (other continuous variable, more than 2 categories, survival analysis etc), but I cannot find much discussion on how to interpret GSEA/GSVA results when they differ or if there are some situations for the scenario I have described (comparison of geneset enrichment in 2 groups) where one method performs better than the other. Is one method better when there are fewer samples or does one method produce more false positives than the other?

I also know there are other single sample enrichment scoring algorithms (singscore for bulk RNA-Seq, and a host of scRNA-seq methods), but I ask about GSVA as it still seems to be a popular choice in many papers.

Can someone provide justification for when to use one method over the other or point me towards any papers that explore this topic?

  • $\begingroup$ Do you want a literature search/pointers or some comments on why to apply one or the other? $\endgroup$
    – llrs
    Mar 11 at 15:44
  • $\begingroup$ Thoughts on when to apply on vs the other is more what I'm looking for but if this is an "unsolved" problem any papers on the the topic would be helpful. I can't seem to find much when I search around $\endgroup$ Mar 12 at 23:55


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.