Fig. 2: Prostate cancer inter-individual and intra-tumor heterogeneity.

A Epithelial clusters were subsetted, and dimensionality reduction (UMAP) and clustering analysis were repeated (n = 78,621 cells, 18 samples). Benign and cancer cells clustered separately. B Stacked bar plots showing the proportion of cells for each cluster that are from (left panel) benign or cancer-enriched libraries and (right panel) which subject (1-10). C UMAP indicating cells from inferred copy number variation (InferCNV) analysis harboring cancer-associated mutations by subject. D Heatmap from inferCNV analysis of cancer cells from Subject 2 indicating regions with inferred CNV gain (red) and loss (blue). Cancer subgroup A shows 10q loss, while the remaining subgroups show 10q is intact, suggesting heterogeneous loss of PTEN (black box). E Example of intra-tumor heterogeneity from Subject 2. Immunohistochemical for frozen sections for PIN4, ERG, and PTEN staining shows cancer cells in top panels are negative for ERG and positive for PTEN. In contrast, lower panels show another group of cancer cells that are ERG-positive but PTEN-negative. Staining was performed across 39 tissue punches, with five specifically from Subject 2. For 2x magnification, scale bar indicates 2 mm. For 10x magnification, scale bar indicates 200 μM. F Heatmap of normalized enrichment scores (NES) from gene set enrichment analysis (GSEA) of Hallmark collection and “Dang MYC targets up gene set” comparing each cancer cluster with the luminal cluster. G Plot showing NES of top 20 pathways by adjusted p-value, comparing aggregated cancer clusters with the luminal cluster by GSEA. Statistics were derived using the fgsea implementation of a two-tailed GSEA. The adjusted p-value statistic is derived from multiple hypothesis testing using Benjamini-Hochberg procedure across all gene sets considered.