GSEA preranked list, values with same rank?

Question

[janders8]

Hi! I have a preranked list of genes that I'd like to run through GSEA. These are genes that correlate with a specific antibody responses, and are results that I have received from our bioinformatician (Spearman correlations). My problem is that the actual correlation coefficients for the genes are almost all the exact same number (either positive or negative)! How can this be and how will this affect how GSEA ranks the genes? Thanks!

Answer 1

[Jean-Karim Heriche]

GSEA deals with ties by ranking them randomly so this may be a problem if you have many ties. See this page on how to use GSEA with a preranked list. I would also be suspicious of all correlation values being the same because this indicates you don't discriminate very well and the ranking may not be robust.

[janders8]

Ok, that was my understanding, too -- thanks for the help!

Answer 2

[Ming Tommy Tang]

I wrote a small post on GSEA http://crazyhottommy.blogspot.com/2016/08/gene-set-enrichment-analysis-gsea.html