Query about NES and FDR q-value

[Dr. Nupur Nigam]

Hi,

I’m running a GSEA analysis on ranked gene list. 

When I am running the GSEA preranked with the hallmarks gene set. 

The GSEA plots are showing an expected curve and NES but the FDR q-value is high. 

All GSEA parameters are kept as default. 

Can you suggest me what could be done better to get a better FDR q-value ? 

Do I need to have a larger number of genes in the rank list? Right now I have 2500 genes. 

Thanks,

Nupur 

...................................................................

Dr. Nupur Nigam, PhD

Center for Cancer Research

National Cancer Institute

National Institutes of Health, 

Bethesda, MD, 20892

Email: dr.nup...@gmail.com

LinkedIn url: linkedin.com/in/nupur-nigam-782280176

[Anthony Castanza]

All forms of GSEA including GSEA Preranked expect information for all expressed genes. Running GSEA with fewer genes than this can adversely affect the results.

Unfortunately if you're running in GSEA Preranked mode already, other than providing information for all expressed genes, the only option you have is to adjust the number of permutations which is unlikely to have large effects on the FDR.

-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 visit https://groups.google.com/d/msgid/gsea-help/CAOE4XhmqfHwKK3tEfQsVEQgJD6OXsBrJBBU9np_jb2v_VgzSfQ%40mail.gmail.com.

[Theo]

Hi Anthony,
If I am not mistaken though the genes with no or very few reads, something like sum<10 are excluded from the gsea analysis

To view this discussion visit https://groups.google.com/d/msgid/gsea-help/CAGCeyZy2AAVOi%2BHzzWbnxoRN4UKheBg0GPfRJcf-GmT43bVZjA%40mail.gmail.com.

[Anthony Castanza]

Hi Theo,

GSEA does not do this automatically, no, although it is probably something we should implement. It does apply a correction factor for genes with low variance.

-Anthony

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