GSEA Output File Changes when changing the plot filter

[Deena Maurer]

Hi, 

I am currently using the GSEA software on my computer (version 4.3.1, Java version 8). I noticed that the rank order of pathways (NES and p.values as well) change for an analysis when I only change the number of plots to graph for each phenotype. All other parameters for the analysis are the same (i.e. input files, phenotype labels, metrics for ranking, etc.) are the same. Why would this be the case?

Thanks,

Deena

[Anthony Castanza]

Hi Deena,

This is due to the way in which GSEA generates the null distribution to perform the significance testing. Each time you run GSEA it uses a random seed to initialize creation of the permutation matrix. On the index.html page for your run in the Comments section we print the random seed used for that run. If you take the random seed from that section and supply it to the "seed for permutation" parameter in the advanced fields section of the application GSEA will regenerate the exact distribution used for the previous run (assuming other parameters that affect the null computation are the same).

Generally I would expect that the differing seeds would only result in a small amount of variability in the results. If you're seeing substantial instability in the values this might be an indication that the default number of permutations might not be enough to generate a reasonable null for your dataset (although I can't make a determination there based on just what you've said here).

Let me know if you have any additional questions, or anything here needs additional clarification

-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/d86f5cb2-b178-438a-bda6-ea3f03198ebbn%40googlegroups.com.

[Deena Maurer]

Hi Anthony, 

Great. That makes sense. Yes, the variability between analysis sets was very small. Thank you very much for the clarification!