Fig. 2: sciCSR reconstitutes productive and sterile IgH transcription levels in different B cell subsets.
a, Histograms of scRNA-seq reads for mature human B cells from peripheral blood FACS-sorted into different B cell subsets (data from Stewart et al.). Counts of 5′-AGCT-3′ motif are displayed to indicate the locations of switch regions. Only reads mapped to IgM, IgG1 and IgG2 are shown here; for the complete IgH locus, see Extended Data Fig. 1. Notice the IgH coding sequences are on the minus strand of human chromosome 14; the horizontal axis is depicted in reverse. b, Alignment of simulated reads (using the polyester package), sampled from IgG1, IgA1 and IgE which are mapped to the 5′ C (light shade) or the C exons (dark shade), using either HISAT2 (teal) or STAR (magenta). Reads sampling are biased either to the 3′ (top) or the 5′ (bottom) end of transcripts to mimic typical scRNA-seq library preparation protocols. c, Alignment of polyester-simulated reads sampled from sterile transcripts of a mixture of IgG (left) or IgA (right) subtypes. The heatmap at the top depicts the proportion of sampled reads mapped to the 5′ C region that is informative of indicating sterile IgH. The bar plots depict the proportions of reads aligned to each isotype using HISAT2 (teal) or STAR (magenta). The ground-truth proportions were noted by crosses (X). d, Dotplot depicting productive and sterile transcription level recovered using sciCSR for the Stewart et al. dataset shown in a. Dot size corresponds to the proportion of cells with positive expression while color intensity represents expression level. e, Proportion of simulated reads sampled from the 5′ C, C and VDJ regions (columns), of sterile and productive IgG1 and IgA1 (rows). The differences in the relative positions of these regions in the transcripts lead to variations in the number of reads attributable to each sterile/productive transcript.
