Detecting differential alternative splicing events in scRNA-seq with or without Unique Molecular Identifiers
Fig 1
(A) Exons were first divided into groups based on their isoform origin. Those exons that originate from the same isoform(s) share the same exon-inclusion level and are thus grouped together. Group 1 (blue) and group 2 (green) are informative for alternative splicing, while uninformative group (grey) is not. Within each exon group, exons were further classified to included (inclusion level = ψ) and excluded (inclusion level = 1−ψ) region. Dark blue and dark green represent included regions. Light blue and light green represent excluded regions. (B) Flow chart of exon grouping and informative read grouping algorithm. (C) Observed read counts of scRNA-seq. Gene-level read count matrix (left) and group-level informative read count matrices (right) are summarized from aligned scRNA-seq data in BAM format. (D) Given gene-level read count matrix, cell-specific technical noises and population level gene expressions are quantified using TASC [17]. Given technical parameter estimates from (C), for each alternative spliced exon group, the statistical inference (exon-inclusion level estimation and inclusion level difference testing) on exon-inclusion levels was made based on a hierarchical model that accounts for technical noise.