Extended Data Fig. 2: Desmosome mutation frequencies in multiple melanoma cohorts and scRNA-Seq analysis of desmosome gene expression in skin tissues.

(a) Related to Fig. 1. Counts of mutated tumors per each desmosome gene (colored points) in the largest published melanoma cohort (Conway²², left) are compared to the SKCM-TCGA cohort (right). The set of desmosome genes has an above-expected number of mutations in both cohorts, with significance determined using bootstrap analysis (*, P < 10^–10 Benjamani-Hochberg correction for multiple testing, see Methods). Desmosome subunits with consistently high mutation frequencies in both cohorts include PKP1, PKP2, DSC1, DSC2 and JUP. DSP has high mutation frequency in TCGA but not Conway. DSG1, DSG3 and DSC3 exhibit a low mutation frequency in both cohorts. Loss of these desmosome cadherins may not provide a selective advantage in melanoma due to functional redundancy with alternate cadherin genes. (b) t-distributed stochastic neighbor embedding (tSNE) analysis of approximately 7500 individual cells from zebrafish melanoma cells. Reanalysis of data published previously²⁹. Color indicates inferred cell type. (c) Same tSNE plot as panel (b), colored by the expression of desmosome genes. (d) Uniform Manifold Approximation and Projection (UMAP) analysis of 239,045 individual cells from healthy human skin, with colors representing inferred cell types of interest (undifferentiated keratinocytes, differentiated keratinocytes and melanocytes). Uncolored (gray) dots correspond to other cell types found in skin including fibroblasts, vascular cells, and lymphocytes. (e) Heatmap showing the average expression levels of desmosome genes in melanocytes as well as differentiated and undifferentiated keratinocytes. (f) Grid of UMAPs from (d), each colored by expression of a different desmosome gene.