Hi everyone, I'm planning on integrating a couple of scRNAseq datasets using either Seurat CCA or Harmony. As the data after integration is scaled, would it still be possible to run GSVA on the integrated/batch corrected data? Thank you so much for your advice!

Hi,

The default settings of the gsva() function expect a matrix of continuous values with rows corresponding to features (e.g, genes, transcripts) and columns corresponding to samples (e.g., individuals, cells, etc.), and a collection of feature sets. Because the GSVA method is non-parametric it should work fine with that kind of input, but it will not make a specific treatment of zeroes in the input matrix, so you may want to filter out lowly-expressed features prior to running GSVA.

cheers,

robert.