Extended Data Fig. 3: Characterizing different epithelial cell derived programs.
a, tSNE view of 23,674 epithelial cells, colour-coded by sample. b, Smoothened distribution for four prostate cancer related signatures in epithelial cell clusters as defined in Fig. 2a. P-values for cluster 10 compared to all other cells for Luminal A, Luminal B are <2.2 × 10−16; CLESs are 0.26 and 0.14, respectively; P-values for cluster 12 compared to all other cells for Luminal A, Luminal B, Hypoxia and PCS1 are <2.2 × 10−16: CLESs are 0.04, 0.79, 0.88 and 0.995, respectively. P values are two-sided and not adjusted for multiple comparisons. c, Smoothened distribution of the indicated signature score, TME and epithelial cells grouped by their assigned cell types and clusters, respectively. P-values (Mann-Whitney U test, two-sided) for cluster 12 CellCycle and cluster 10 basal/intermediate signatures compared to all the rest cells are < 2.2 × 10−16; CLESs are 0.998 and 0.98, respectively. d, GO terms enriched in CellCycle subtype (left) and contingency table showing number of cells in G2/M for CellCycle subtype compared to all the other cells (right), One-sided P value calculated with Fisher’s exact test. OR = 5.7. e, Comparison of BCR-free rate between the high and low groups stratified using CellCycle signature across multiple datasets. P-values are calculated using Cox proportional hazard model (CoxPH) and not adjusted for multiple comparisons. Numbers in brackets show 95% CI for PH.
