Hello all, I'm working on single-cell RNA-seq data and trying to using seurat for finding marker gene on cluster. I found that I need to run
PrepSCTFindMarkers() before applying
FindAllMarkers() on non-integrated "corrected" count derived by SCTransform. However, I cannot find any documentation about the reason for that. If anyone of you know the reason behind that?
Moreover, I'm thinking of using non-integrated pearson residual instead the "corrected" count from SCT. But my concern is that pearson residual aims to reduce the variance in expression across cell and individual, so I'm wondering if the sources of biological difference across clusters are reserved for DE testing?
Thanks for any help.