Question about using scMerge with Seurat workflow

[yls2g13]

Thanks for a great package @YingxinLin and @kevinwang09. I'm considering setting this up for a Seurat workflow, and need to subset CD8+ T cells from a CD45+ Total Immune Capture. I understand scMerge generates a normalized matrix as a result.

Can I confirm that after subsetting for CD8+ cells, if re-normalization is not needed? My current workflow looks like this:

# subset cells obj from `seu`, my total immune capture object
# My RNA assay has the normalized assay `assay(merged_sce, "scMerge2")` from scMerge 
cd8 <- colnames(seu)[grep("T cells", seu$High.hierarchy.cell.types, invert=TRUE)]
seu <- subset(seu, cells = cd8, invert = TRUE) 

# Normalization and scaling
# # so <- NormalizeData(so) # no normalization
so <- FindVariableFeatures(so)
so <- ScaleData(so, vars.to.regress = vars.to.regress)
  
# Dimensionality reduction
so <- RunPCA(so, npcs = 50, verbose = F)
so <- RunUMAP(so, dims = 1:npcs, verbose = F)

[SOMS-Maintainer]

Yes, it is one of few batch effect adjustment software that calculates adjusted per-gene numbers. After subsetting, re-normalization is not needed. Was your data collected over a long time? Can you actually see batch effects in it that needs handling?

[yls2g13]

Thanks for your quick response. Is the workflow I shared above accurate?
Our data was collected over a long time. I'm also sharing a UMAP here coloured by batch, there's quite a bit of separation.

[SOMS-Maintainer]

Yes, that's fine to do.