I’m trying to run the latest version of DecoupleR-py (v2.0.1) now implemented into the scverse.
I followed the old tutorials of DecoupleR-py before and seem to remember them using the raw counts of an adata to perform the enrichment analyses such as with mt.ulm for example. However now in the new tutorial docs, the enrichment analyses are run on the log-normalized counts of an example adata. Yet I see that the ‘use_raw’ argument is still an option in the functions.
I was just wondering whether you would recommend running enrichment analyses in DecoupleR on raw or normalized counts?
New user of DecoupleR here - wanted to also follow up on this topic.
I have been trying to follow the DecoupleR Pseudobulk Enrichment Analysis tutorial. The tutorial does not specify whether the raw counts or some normalized counts are used from the pseudobulk, and there is no explicit setting or invocation of the .X matrix. We tried running exactly what was in the tutorial docs and ended up with differential expression results that had almost exclusively negative log2 fold change values. We wanted to know whether this was an actual result or possibly an artifact of using the wrong matrix for counts.