Single cell DEG analysis. Wildly different results from different FindMarkers test
Hello and thank you in advance for your help.
I am doing a single cell RNA seq experiment and analyzing data with Seurat. I have four treatment groups, which I integrated, then identified clusters of different types of cells in the brain. Everything looked exactly right at this point. Then I wanted to test for DEGs within each cluster. Using the standard Seurat workflow of switching default assay to "RNA" and then using FindMarkers(), I get kind of strange results. Here is a sample volcano plot from one cluster:
If I switch to a statistical method in FindMarkers (like negbinom) that uses RNA counts instead of scaled data, things look much more as I would expect. Here is the same cluster, tested using counts as data.
I have triple checked that I am performing the Seurat workflow as per the tutorial here:
It just seems so odd to me that when I do DEG testing the apparently correct way, I get huge LogFC's and also a weird looking volcano plot with many many significant (padj < .05) DEGs with logFC close to zero.
Has anyone encountered a similar issue? Any help would be great. Thank you!
• 1.5k views