Pathway analysis with co-variates for RNA-seq data

Question

[JJ]

Hi all,

I was wondering if someone knows of a tool to perform a pathway analysis with co-variates for RNA-seq data.

What I mean in more detail is the following: I have two groups (cond. 1 vs. cond. 2) but I also have co-variates which will have an high impact on the expression (20-30 samples in total, balanced design). Hence I am not sure that a standard DE analysis e.g. with voom/limma with co-variates will lead to significant results. I more expect that the same pathways will be up or down regulated than particular genes. Hence, I am looking for a gene-set based analysis where I can also give co-variates as parameters.

Thanks for your help!

Answer 1

[Kevin Blighe]

With GSVA, you can enrich your dataset against MSigDB, KEGG, Reactome, and Biocarta in an unbiased fashion and then perform a differential pathway / signature analysis while adjusting for covariates (usually via limma). GSVA will essentially return to you your same dataset with your genes replaced by pathways and their respective levels of activation.

I would highly recommend getting to grips with the program as it is very powerful for a diverse range of analyses. To give you an idea of things for which I have used GSVA:

• simple pathway / signature enrichment
• differential pathway / signature analysis while adjusting for covariates
• disease predictive modeling using pathways, not genes
• inferring scRNA-seq cell-types in bulk data (akin to cellular deconvolution)

Kevin

[JJ]

thanks so much for this pointer!

[Kevin Blighe]

If you need help with code, then give me a shout.

[JJ]

will do - thanks so much!