Difference in DGE and GSEA ?

[Pierre-Alex Laval]

Hello again,

I have several question regarding the DGE and GSEA data available from the xenabrowser.

From what I gather, once you have 2 different groups you can make a DGE or a GSEA from those 2 groups to see the most dysregulated expresision and signature. 

I make 2 groups from my data, EIF2AK4 (GCN2) low vs high. 
I performed a GSEA and a DGE analysis from those two groups comparing low versus high (so First group is low, second is high).

-Looking at the DGE, the most dysregulated gene is EIF2AK4 (which is normal), but looking at the Characteristic_direction DGE performed by the GSEA pipeline, the most dysregulated gene is KCNEV1, EIF2AK4 is a bit higher. I don't know if that's normal ? Or in general what's the difference between DGE and CD_DGE ? 

-For the GSEA analysis, is there a way to know which comparison is done for the signature ? I assume it compares the first group to the second, and therefore if NES negative your signature is underrepresented in the first group compared to the second but I want to be sure.

Thank you !

PA

 

[Mary Goldman]

Hi PA,

For both the Differential Gene Expression and the GSEA analyses, there are two options for generating the Differential Gene Expression Analysis/Signature: limma voom and characteristic direction. By default we pick limma voom, just so there is a default to begin with, but you may prefer characteristic direction. I recommend reading more about both before deciding which is best for your analysis.

Here are the papers that are cited at the bottom of the analysis pages: 

limma voom: Ritchie, Matthew E., et al. "limma powers differential expression analyses for RNA-sequencing and microarray studies." Nucleic acids research 43.7 (2015): e47-e47.

characteristic direction: Clark, Neil R., et al. "The characteristic direction: a geometrical approach to identify differentially expressed genes." BMC bioinformatics 15.1 (2014): 79.

Yes, you are correct about the direction of the signature for the GSEA analysis.

Best,

Mary

-----

Mary Goldman (she/her), Design and Outreach Engineer 

UCSC Xena

UC Santa Cruz Genomics Institute

Revealing life's code

--
You received this message because you are subscribed to the Google Groups "UCSC Xena and Cancer Genomics Browser" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ucsc-cancer-genomics...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ucsc-cancer-genomics-browser/06903b23-57f0-44e1-a68e-e48036f12500n%40googlegroups.com.