Perform GSEA without gene expression data

Question

[Eliza]

Hi ,

I was wondering if it is possible to perform the GSEA https://www.pnas.org/doi/epdf/10.1073/pnas.0506580102 without having gene expression data ?

Thank you

Answer 1

[rfran010]

No. GSEA needs expression values to rank gene enrichment between conditions. If you have gene lists, you can look into gene ontology tools which only require gene names or IDs.

You can also input your gene list here and select different gene sets to compute overlaps, probably as close as you will get to using the GSEA tool without expression values. https://www.gsea-msigdb.org/gsea/msigdb/human/annotate.jsp (there is a mouse collection as well. The menu on the left, Mouse Collections > Investigate)

[Eliza]

@rfran010 Thank you as I understad in here https://www.gsea-msigdb.org/gsea/msigdb/human/annotate.jsp i can compute overlaps between my gene set and gene sets in MSigDB, is there a tool that perhaps an compute the overlap beween a genes from my experiment -L and predefined genes -S as described here :

[rfran010]

What gene set do you have?

In general, I only say gene set when I refer to a predefined gene set (S). For example, imaginary gene set for Neutrophil Degranulation has GeneA, GeneB, GeneC, ... Gene M.

L is the list of your own genes. In these case of GSEA, it takes L sorted by enrichment (using expression values) in Condition1 vs Condition2. So, here L needs to contain all detected genes and GSEA automatically determines if a subset of those genes are enriched in Condition1 or Condition2. So if a majority of GenesA...M are highly upregulated in Condition2, then Condition2 would be enriched for Neutrophil Degranulation.

In the link, Compute Overlaps does a similar enrichment analysis, but you have to manually subset your list L. Presumably, you will have a list of genes that are enriched in one condition or another. Maybe, a list of upregulated genes in Condition2. You can take this list and copy it into the link, then select any gene set category you are interested in.

In my example, I have a list of the top 500 enriched genes after depletion of my targeted protein. I am interested in the Gene Ontology sets and Cell Type Signature sets, so I have selected those options. I then click "compute overlaps" and see that Apoptotic Processes is one of the most significantly overlapped gene sets with my 500 upregulated genes. This makes sense since in my experiment I know all my cells will die within 12 h after the timepoint the experiment was performed. I could then repeat process for downregulated genes, or maybe genes upregulated uniquely in my experiment compared to a different protein depletion, or any other sort of logical subset.

I know this isn't a succinct answer, but I am not too clear on the question.

[Eliza]

@rfran010, so i have the following data: I'm conducting research; in this research i have 200 genes that have a pvalue <0.05 ( which i presume can cause a severe disease instead of mild in the disease im researching) which i want to research further -> the L list of my own genes . also the the S list of predefined genes associated with Epidermolysis bullosa disease . and i want to know if the gened from the Epidermolysis bullosa are overrepresented in my L list . so i whished to understand if i could do GSEA with out having the expression data on L ,thank you :)