GSEA vs Functional Annotation

Question

[turcoa1]

Hi there,

I just finished a differential gene expression analysis with DESeq2. I now have a list of the top 500 most downregulated genes and want to annotate these genes to see what they are involved in. I have been looking into gene set enrichment analysis but I am confused in regards to the statistical test employed in GSEA. If I am already working with genes that I know are under expressed in a single condition, why would I need to compare differences in these conditions again? I am trying to understand how I can simply visualize the gene sets without employing a statistical test as I only want to focus on genes downregulated in one condition.

[Ming Tommy Tang]

read a thread here

Pathway or gene set enrichment analysis is frequently used in genomic studies. Make sure you understand it with these 8 resources: 👇 🧵

— Ming "Tommy" Tang (@tangming2005) June 22, 2023

Answer 1

[LauferVA]

What you have done with DESeq2 is to generate results for individual genes.

By contrast, what GSEA tells you is, if, taken together, all those individual gene results also constitute a significant difference to the entire pathway.

DESeq2 Result : DE gene : : GSEA : DE pathway

Because GSEA essentially is a framework of rigorous statistical analyses, one cannot perform GSEA "without employing a statistical test".

It is true you can visualize gene sets without generating differentially expressed pathway results. However, two caveats occur to me:

1) you would not be able to comment on whether the pathway is significantly dys-regulated without GSEA or another like methodology, you'd just be looking at the genes laid out in some fashion.

2) most well-developed software packages used to visualize gene sets also rely on the results of GSEA or some algorithm like GSEA (see, for instance, enrichplot). As such, if you wish to look at gene sets, even then I'd probably still run GSEA.