Figure 1

Linking mutation signatures to exposures or endogenous mutational processes. One can either (a) focus on signatures of one mutagen at a time or (b) study mixtures of signatures. One can study signatures of single mutagens either (c) via experimental approaches, or (d) via observation of mutation signatures in the exome or genome sequence of tumors with a known predominant mutational exposure. Some tumor exomes harbor only a handful of somatic point mutations, and presumably these tumors arise from causes other than mutagenesis. For many cancers, typical numbers of somatic point mutations in exomes are 60 to 300 [9–12]. Highly mutated cancers sometimes have >3,500 mutations per exome [13]. Typical numbers for genomes of cancers such as those of the lung or stomach are >15,000 [14, 15], and a few highly mutated genomes harbor >400,000 somatic point mutations [16]. Among experimental approaches, one can use (e) reporter constructs and observe mutations in short sequences. This allows inference of relatively simple signatures, for example (g) signatures involving only single nucleotide mutations. (f) By sequencing the exome, or, ideally, the genome of a mutagen-exposed, clonal cell line, one can (h) infer a more informative, extended signature, for example one that includes the trinucleotide contexts of single-nucleotide mutations. One can also infer extended signatures by sequencing the exomes or genomes of tumors with known, predominant exposures (d). Extraction of mutation signatures from mixtures of signatures (b) requires somatic mutation catalogs from the exomes or genomes of large numbers of tumors. The most recent studies have looked at thousands of catalogs. (i) Procedures based on NMF (non-negative matrix factorization) allow (j) simultaneous inference of a set of extended mutation signatures and the contributions of each inferred signature to each tumor's mutations. (k) Extended signatures derived from mixtures of signatures (j) can be matched to extended signatures that were experimentally determined or inferred from known predominant exposures (h), thereby providing information on (l) exposures that contributed to tumors with mixtures of mutation signatures. (m) Alternatively, extended signatures extracted from mixtures can be correlated with information on mutagenic exposures or on endogenous mutagenic factors, allowing inference of the causes of mutation signatures (l). (n) The causes of some signatures will remain unknown and require further research.