RNA-seq Heatmaps based off pathways

Question

[josh.faust]

Hey all!

Very new to RNA-seq data analysis and have been using normalized gene counts (TMM) to do subsequent analysis based off of log2FC and p-value etc. I was wondering if anyone had any recommendations on doing pathway analysis in terms of generating a heatmap. I have used pathview and was able to get some data out of that, but I want to generate heatmaps based off the pathways to show how the expression pattern of genes associated in the pathways change. Is there any easy way to do this? I can only think about pulling out select genes that are predetermined to be in the pathway, but I don't even know if there is a resource that would have all this information.

Any advice would be useful!

Thanks :)

Answer 1

[jared.andrews07]

You seem to be looking for Gene Set Enrichment Analysis (GSEA). This can be done a few different ways, either with your raw counts, or more easily with a pre-ranked gene list. Often the ranking is done by the signed test statistic. fgsea is a very easy R package to use for the pre-ranked style. This is useful for identifying pathways/gene signatures that differ between two conditions or groups.

There is also single sample GSEA (ssGSEA), which provides a score for each sample. The GSVA R package includes this functionality and several other methods for generating sample-specific pathway scores. I am actually partial to UCell, as it provides a nice 0 to 1 scale that is comparable across samples. It was designed for single cell RNA-seq, but works well for bulk as well. You can make heatmaps with the scores returned from pretty much any of these methods.

If you use any of these methods, you can easily add the scores to the samples dataframe of your DGEList, and then use dittoSeq to very easily make the heatmap for any pathways of interest.