SCTransform compatibility with scCATCH
bounlu opened this issue · 3 comments
The scCATCH document says use log1p normalization as the default method of Normalizedata function in Seurat. However, in later versions of Seurat, SCTransform is integrated and suggested which replaces several steps including the normalization. Does it still use the same method of log1p as suggested? Is such normalized Seurat object still safe and good to use with scCATCH?
After SCTransform workflow, there still exists an assay called 'RNA'. You can use NormalizeData function in Seurat on this 'RNA' assay and the log1p data is stored in the 'data' slot under 'RNA' assay.
Indeed I proceeded with SCT assay and it still worked fine in my case. However, findallmarkers from scCATCH didn't work fine as later not all my clusters were assigned to cell types, however FindAllMarkers function from Seurat worked great and scCATCH could only after that find cell types for all clusters.
Hello, findmarkergenes is to find the specifically highly expressed genes for each cluster which is more strict than the FindAllMarkers. So, some clusters may be assigned with no cell types as no marker genes found when there are lots of clusters. Under this condition, FindAllMarkers is a good choice. findmarkergenes is more proper for scRNA-seq datasets with limited clusters.