There are ties in the preranked stats (3.06% of the list).

[Gordon Zheng]

Is there a way to combat this error? I am following the guide set by https://stephenturner.github.io/deseq-to-fgsea/

using FGSEA with eps = 0 as oppsed to nperm=1000,

I also included the code: mutate(rank = rank (log2FoldChange, ties.method =  "random"))

in attempt to resolve the issue by no success. Any suggestions? 

[Anthony Castanza]

Hi Gordon.

Unfortunately fgsea is a 3rd party reimplementation of the GSEA algorithm (with some changes to the significance testing) that we don't support or endorse. This is only the support forum for the 1st party, original, implementations of GSEA.

If the random method is failing to break ties, I can only really suggest reaching out to the developers of fgsea directly.

That said, you might try using a different piece of the DESeq2 dataset as your ranking metric. I've had good results with the wald statistic that it produces. If that also doesn't work, you can try manually computing a scaled log2fc using the formula log2(fc)*-log10(pValue). This computation may provide some additional tiebreaking and has the additional advantage of incorporating both magnitude and significance of the changes into the ranking.

I don't know if fgsea supports these as direct input however, so I might direct you to the GSEA Preranked method within the standard GSEA desktop application that we provide

-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/d765a427-fcd7-45e3-8b97-0fb7d6f9fa86n%40googlegroups.com.