Fig. 5
Oxidative damage patterns are reflected in cancer mutagenesis. a Boxplots of the proportion of C-to-A SNVs (including the reverse complement G-to-T) in genomic regions of high GC content (> 50%). Tumor samples are separated into four groups according to Mutational Signature 18 contributions (n< 0.1 = 1398, n0.1–0.4 = 322, n0.4–0.6 = 540, n> 0.6 = 141). Asterisks indicate significance of p < 0.001 by Wilcoxon rank test comparing the different Signature 18 proportions to Signature 18 < 0.1. Bar plots depict the original COSMIC mutational signatures. Tumors that are high in Signature 18 display lower proportions of SNVs in GC-rich regions, while tumors with mutations in OGG1, APEX1, or FEN1 show higher proportions. b Boxplots of the proportion of T-to-G SNVs (including the reverse complement A-to-C) in genomic regions of high GC content (> 50%). Tumor samples are separated into four groups according to Mutational Signature 17 contributions (n< 0.1 = 2255, n0.1–0.25 = 78, n0.25–0.5 = 59, n> 0.5 = 9). Asterisks indicate significance of p < 0.001 by Wilcoxon rank test comparing the different Signature 17 proportions to Signature 17 < 0.1. Tumors that are high in oxidative damage signatures display lower proportions of their respective signature mutations C-to-A or T-to-G in GC-rich regions. c Boxplots of the proportion of C-to-A SNVs in genomic regions of high GC content (> 50%). Tumor samples are separated into those that are Pol E-proofreading defective (n = 8) and to all other tumors (n = 2694). Asterisks indicate significance of p < 0.001 by Wilcoxon rank test comparing the PolE proofreading deficient to competent. Tumors that are proofreading defective and high in Signature 18 display lower proportions of SNVs in GC-rich regions. d Metaprofile of SNV rates over ~ 23,000 protein-coding genes in proofreading-defective and control tumors. The damage profile is overlaid for comparison. The oxidative damage-dependent SNV profiles in proofreading-defective tumors show similar distributions to AP-sites, whereas the pattern is lost in control tumors. e–h Metaprofiles of SNV rates centered on CTCF binding sites (n = 48,671; e), transcription factor-binding sites in DHS regions (n = 253,613; f), CpG islands (n = 27,443; g), and G-quadruplex structures (n = 359,449; h). SNV profiles in proofreading-defective tumors mimic the damage profiles. i, j Metaprofiles across 848,350 Alu (i) and 2,533 LINE elements (j). SNV rates in proofreading-defective tumors are reduced compared with damage profiles