rcastelo
Support for GSEA plots
Would it be possible to implement a function that can plot (for a given gene set) a GSEA plot similar to the one here:
Hi, sure, I guess you mean to show such a plot for a specific column from your input data matrix, right?
In the meantime, you can have a try at the interactive web app described in the vignette, and you could try to obtain the plot from there.
I tried that actually but the web app couldn't run GSVA because it required more memory, I guess, I'm not sure. Normally I run GSVA using BPPARAM = BiocParallel::MulticoreParam(workers = 100) on an RStudio Pro server with 128 cores and tons of memory. Even then it takes a while with the data and gene set list I am using. It would be nice if I did not have to rely on the shiny app.
I suggest we don't discuss the web app issue here as it is another issue. I can open up one for it, if you like.
I see, we will provide an alternative solution, no need to discuss the web app problem here. What kind of data (single-cell, bulk RNA-seq, proteomics, etc) and dimensions (rows/genes/transcripts x columns/cells/samples) do you need to input to GSVA?
Around 50,000 genes x 4,000 samples as the expression data, and 20,000 gene sets.
ok, thanks, is the data sparse (more than 80% of the values) like if they were single-cell data?
Not sparse at all. We already prefilter the data to remove genes with constant or no expression, so there is a TPM value for each gene in each sample.
Dear @cihanerkut yesterday we released a new version 2.0 of GSVA, see https://bioconductor.org/install for installation instructions, that includes a new function gsvaEnrichment() that will allow you to make GSEA-like plots from GSVA results. Please take a look to the help page of gsvaEnrichment() and do not hesitate to write back if anything is unclear.
Dear @rcastelo , thanks a lot for your great support. I look forward to trying it :)