Figure 4.

Selection pressure against damaging mutations

(A) The ratio of the number of nonsynonymous mutations to synonymous mutations. We added 1 to the denominator values to avoid potential 0.

(B, E, H, and K) The observed ratios of the numbers of nonsynonymous, frameshifting, stop-gain, and stop-loss mutations to the numbers of synonymous mutations vs. their corresponding ratios expected by chance alone.

(C, F, and I) The numbers of frameshifting, stop-gain, and stop-loss mutations per cell.

(D, G, and J) The numbers of observed frameshifting, stop-gain, and stop-loss mutations vs. their corresponding numbers expected by chance alone (in log2-transformed ratios).

To estimate the number of mutations expected by chance alone, we first used SigProfilerSimulator³¹ to randomly generate the same number of SNVs and INDELs as the observed numbers with also the same mutation signature, then annotated the artificial mutations with ANNOVAR³⁶ to determine the number of mutations in each functional category, and finally repeated the above two steps 2,000 times to reach stable estimations. Each dot represents a cell. The p values in (A), (B), (C), (D), (F), and (H) were estimated using linear mixed-effects models, two-tailed. In (E), (G), and (I), ns represents p > 0.05 and ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001, separately, which were estimated using binomial tests, two-tailed. Boxplot elements are defined as follows: center line indicates median, box limits indicate upper and lower quartiles, and whiskers indicate 1.5× interquartile range.