Fig. 4

Annotation of CDSeq-identified cell types. CDSeqR was run with the Reduce-Recover option using 800 genes per block and 10 blocks. A reference scRNA-seq dataset was provided for cell type annotation using clustering analysis (a and b) and correlation analysis (c and d). a UMAP embeddings of reference scRNA-seq data (color dots) and pseudo single-cell data generated from the CDSeq-estimated cell-type specific gene expression profiles (color squares). Each distinct color in the UMAP plot denotes a cell type provided with the scRNA-seq annotation given in [20]. There are 7 major cell types provided with the scRNA-seq: excitatory neuron (Ex), inhibitory neuron (In), oligodendrocyte (Oli), microglia (Mic), endothelial (Endo), astrocyte (Ast) and oligodendrocyte progenitor cell (OPC). The deconvolution result used 7 as the number of cell types and for each CDSeq-estimated cell type, it generated 10 pseudo cells which correspond to a color square in the UMAP plot. The colors of the squares (i.e. CDSeq estimated cell types) were determined based on its annotation result. b Correlations between the resulting CDSeqR estimated cell type proportions and true proportions; c correlations between the CDSeqR estimated cell-type-specific GEPs and scRNA-seq GEPs; b correlation of CDSeq-estimated cell type proportions with true proportions using correlation analysis (see Methods for details). When reference for cell annotation is not provided, CDSeq will perform clustering analysis on the pseudo single-cell data generated from the CDSeq-estimated cell-type specific gene expression profiles and perform differential gene expression analysis to identify the marker genes in each cluster using the Seurat software [24, 25]. e Expression level of PLP1 (Oligodendrocyte marker [41]) in the clusters of CDSeq-estimated cell types (cell type number is 7); f Expression level of PLP1 in human brain prefrontal cortex scRNA-seq clusters [20]. See Additional file 1: Figs. S10–S20 for more details