Normalization and Cell Cycle scoring

Code

# Enrique Blanco Carmona
# e.blancocarmona@kitz-heidelberg.de
# PhD Student – Clinical Bioinformatics
# Division of Pediatric Neurooncology (B062)
# DKFZ-KiTZ | Germany

#--------------------------------------------------------------------
# 4 - NORMALIZATION USING SCTRANSFORM AND CELL CYCLE SCORING
#--------------------------------------------------------------------

# In this process, we want to regress out the effect of 5 variables:
# - nCount_RNA: UMIs. We observe different numbers across samples.
# - nFeature_RNA: genes. We observe different numbers across samples.
# - percent.mt: % of mitochondrial RNA in the cells.
# - S.Score and G2M.Score: Cell cycle effect. We want to be sure this is not a bias in the data.

# Get S phase genes.
s.genes <- Seurat::cc.genes.updated.2019$s.genes
# Get G2-M phase genes.
g2m.genes <- Seurat::cc.genes.updated.2019$g2m.genes


# We normalize first using SCTransform and removing the effect of UMIs, genes and percent.mt.
# Process followed as in https://github.com/satijalab/seurat/issues/1679#issuecomment-557781838

sample <- Seurat::SCTransform(sample,
                              assay = "RNA",
                              new.assay.name = "SCT",
                              vars.to.regress = c("nCount_RNA", "nFeature_RNA", "percent.mt"))


# With the data normalized, we asses the cell cycle effect.
sample <- Seurat::CellCycleScoring(object = sample,
                                   s.features = s.genes,
                                   g2m.features = g2m.genes,
                                   set.ident = FALSE)


# And then, since now S.Score and G2M.Score are stored as metadata variables, we can normalize again using also these variables in the regress out.

sample <- Seurat::SCTransform(sample,
                              assay = "RNA",
                              new.assay.name = "SCT",
                              vars.to.regress = c("nCount_RNA", "nFeature_RNA", "percent.mt", "S.Score", "G2M.Score"))
# Please note that this second normalization uses the data in the "RNA" assay, this is, the unnormalized data.
# Therefore, the data we will work on is only normalized once.