Extended Data Fig. 7: Gene programs, and cell cycle changes impacted by KRAS variants.

a-d. Gene programs impacted by variant classes. a-b. UMAP embedding of single-cell profiles (b), colored by program scores (color bar) and labeled by selected Gene Ontology biological processes enriched in genes from each program (a). c. CDF for program scores (x axis) for each variant, colored by class. d. Average expression (z-score, color bar) in cells of each variant (columns) of genes (rows) most correlated with the mean of the expression program. e. Difference (dot color) in mean expression of each gene program (rows) between the cells in each cluster (columns, as in Fig. 4e) and all other cells, and the significance of this difference (dot size, -log10(adj. p-value, Benjamini-Hochberg procedure), Kolmogorov-Smirnov test, Methods). Colored border: BH FDR<10%. f. ROC curve of the true positive (y axis) and false positive (x axis) rate when using each PC (color) to distinguish between single cells with synonymous variants and those with variants in hotspot positions 12, 13 and 61. Color legend: Area Under the ROC curve (AUROC) for each variant. g,h. Principal component analysis. g. UMAP embedding of single-cell profiles, colored by principal component (PC) scores (color bar), for each of the first 10 PCs. h. CDFs for the PC scores (x axis) for the cells of each variant, colored by class. i. Variant expression (y axis) as a function of PC 3 scores (x axis), for each variant (dots). Variants are colored by variant class. j. Proportion of cells in each cell cycle phase (y axis) among cells carrying each of 98 variants (dots, n=median 1058 cells/variant) across variant classes (x axis). P-values for a two-sided t-test have been adjusted using the Benjamini-Hochberg procedure.