Fig. 2

A workflow of sc/snRNA-seq data preprocessing. After obtaining the single cell or single nucleus sequencing count data, a series of quality control processes are conducted to filter low quality cells with unusually high or low gene coverage or sequencing depth, unusually high mitochondrial content, ambient RNA, and doublets etc. QCed count data is normalized by either a global scaling approach or advanced parametric modeling of the zero-inflated count data distribution. If there exists batch or condition-specific clustering of the cells, a data integration method like MNN and CCA can be used to correct the batch difference to ensure that cells of the same cell type cluster together