I have a scRNA sample (genes x cells) of peripheral blood with 8 annotated cell clusters (B-cells, T-cells, ..) and want to use gene enrichment / pathway analysis on each cluster of this single sample. I looked at some sc packages that have this function implemented. One calculates the mean gene expression for each cluster and uses this matrix as input for GSVA (genes x cluster). Another package also uses the mean gene expression for each cluster as input for ssGSEA.
Is one or any of this ways correct or is there another way like GSVA of the raw genes x cells matrix and calculate the cluster enrichments from there somehow?
And which one can I use to compare the results of one cluster - for example B-cells - with the B-cell cluster results of a another control subject? GSVA of the raw genes x cells matrix? Only one input value per subject probably won't work with limma. Or is it better to use for example fGSEA with pre-ranked DESeq2 results?
Thanks in advance!
Which packages are these and why do you think either one is not correct?
Why not use GSVA?
Sorry, let me rephrase that: I don't think either of them is not correct but different approaches that give different results. My question is if anyone can argue positive or negative sides of either method in this context.
The second question was about the kind of input for GSVA - either mean gene counts per cluster or raw gene counts per cell and combine the results for euch cluster somehow.
GSVA can take either raw counts or logCPM/logTPM values (see the documentation). If you need the results per cluster, then you can use the average expression per cluster. It is not designed for scRNA-seq.