Extended Data Fig. 7. Further details related to the updated catalog of ID signatures.

a. Spectra of COSMIC ID15 and 16. These two signatures are not discovered from our PCAWG reanalysis with MuSiCal. b. Indel spectra of the three TCGA whole-exome sequenced samples from which COSMIC ID15 and 16 are discovered by the PCAWG consortium⁵. c. Number of indels vs. number of SBSs for all TCGA samples, highlighting that these three samples have exceptionally high indel counts but low SBS counts. d. Variant allele frequency (VAF) distributions of indels and SBSs for these three samples are compared to those for the other TCGA samples. Indels in these three samples have particularly low VAF, suggesting that they are likely artifactual. e. COSMIC ID4 is resolved into multiple signatures in our updated catalog. COSMIC ID4 is decomposed using MuSiCal ID4, 19, and 24 with NNLS, and the corresponding exposures are annotated next to each of the MuSiCal signatures. The reconstructed signature has cosine similarity of 0.996 with COSMIC ID4 and is shown at the very bottom. f. The TOP1-associated ID spectrum observed in RNase-H2-null cells from⁴² (top) is compared to the reconstructed spectra using the COSMIC (middle) and the MuSiCal catalog (bottom). The MuSiCal catalog better reconstructs the experimentally derived TOP1 signature. The TOP1 signature is more similar to MuSiCal ID4 (cosine similarity = 0.87) than COSMIC ID4 (cosine similarity = 0.83). Specifically, COSMIC ID4 contains longer (3- and 4-bp) deletions that are not observed in the TOP1 signature.