(A) Dot and sashimi plots showing LIMCH1 compartment-specific exon skipping (involving 5′ splice site [5′ SS] 41,619,440 and two 3′ splice sites [3′ SS] 41,638,932 and 41,644,500) impacting a protein domain of unknown function DUF4757 (shown by the purple color on the gene structure) across cell types and 10X and SS2 data from both human individuals. Each dot shows junction expression for the splice junction from the 5′ SS to one of the 3′ SS’s, with dot size proportional to the fraction of junctional reads supporting the splice junction in that cell type and dataset. Columns of dots are biological replicates; the first column is the individual 1 10X dataset (circles) and the next two columns are SS2 datasets from individuals 1 and 2 (squares). Cell types are grouped in two sets depending on the sign of the median SpliZ score in 10X data from human individual 1. The thickness of the sashimi arcs represents the fraction of the reads mapping to each 3′ SS when all datasets and corresponding cell types for the sashimi arc are grouped together. The box plot for each cell type shows the distribution of the weighted average 3′ SS (weights being the number of reads aligning to each 3′ SS in the cell) for each cell and the reads are assigned 1 (for those aligning to the closer 3′ SS) and 2 (for those aligning to the farther 3′ SS). Stromal cells including vasculature smooth muscle cells and fibroblasts always include the exon (higher fraction of reads aligning to the splice site at 41,638,932), while epithelial cells including bladder urothelial cells skip with >50% rate. (B) Unsupervised k means clustering results in 78, 84, and 95% accuracy of compartment classification for the stromal, epithelial, and immune compartments, respectively, for individual 1, and 70, 100, and 49%, respectively, for individual 2. (C) The SpliZ scores of the genes ATP5F1C and RPS24 together separate compartments in both human individuals. Each dot represents the SpliZ score in a single cell and is color coded by the compartment. (D) Using the spliced read counts for each gene rather than the SpliZ does not separate the compartments, showing that this separation is not driven by gene expression differences. Each dot represents the number of spliced reads in a single cell and is color coded by the compartment.
© 2021, Olivieri et al. This figure was also included as Figure 2D in Olivieri et al., 2021, published under the Creative Commons Attribution Non-Commercial No Derivatives 4.0 International License.