Extended Data Fig. 3: Association of RHPs with cell cycle, pool identity and cancer types.

a, Average distribution of cell cycle phases in cells with high RHP scores (top 5%). b, Median RHP scores of cells in each phase of the cell cycle. Cell cycle state was estimated for each individual cell based on the relative expression of the G1/S and G2/M metaprograms (see Extended Data Fig. 2c). For each RHP in (a) and (b) we only considered the respective model cell lines (see Supplementary Table 3). c, Distribution of RHPs (n = 680) across the 9 pools (H corresponds to the custom HNSCC pool). Each RHP was observed in multiple pools, underscoring the lack of pool-specific effects and the robustness of RHPs. d, The fraction of cell lines from each cancer type (rows) observed in each RHP (columns) are indicated by circle size. Red circles depict significant enrichments (P<0.001 by hypergeometric test). Left panel shows the fraction of each cancer type in our dataset and the number of cell lines profiled. e, Hierarchical clustering of cell lines based on their set of RHPs (bottom panel). Clusters composed of more than 5 cell lines are annotated (1–9). Relevant cell line features are shown on top and their associations with each cell line cluster was tested using hypergeometric test. Significant associations (P < 0.001) are indicated by black squares and arrows. These include associations of cluster 2 with melanoma, cluster 4 with HNSCC and squamous cells, and cluster 8 with wild-type p53.