Figure 5

Reconstructing the catalog of somatic mutations in a cancer genome as superimposed mutation signatures at varying levels of exposure. (a) Each signature is represented by one of the bar charts, and consists of the relative proportions of different types of mutations in that signature. For example, in Signature 1, C > T mutations make up almost half of the total number of mutations, whereas T > A mutations constitute only about 10% of the total. (b) Each of the three signatures contributes a different number of mutations to the actual catalog, represented in the 'pie chart'. In this example, Signature 1 contributes 1,000 mutations, Signature 2 contributes 1,500, and Signature 3 contributes 750. The 1,000 mutations from Signature 1 are allocated according to the bar chart that represents the proportions of different types of mutations in this signature. In this case, Signature 1 would contribute approximately 50% × 1,000 = 500C > T mutations. Signature 2 would contribute approximately 9% × 1,500 = 135C > T mutations. Signature 3 would contribute approximately 10% × 750 = 75C > T mutations. The total number of C > T mutations in the reconstructed catalog would be 500 + 135 + 75 = 710. The reconstruction of the (d) actual catalog is approximate, and in this example, the reconstruction does not account for 65 mutations, approximately 2% of the total in the actual mutation catalog - the gray noisy line in (c). This figure is a simplification; in fact, in references [8, 75, 76], signatures are composed of nucleotide mutations in their trinucleotide contexts, as shown in Figures 3c,d,h and 4a-d. The mathematical procedures for approximating observed catalogs from mixtures of trinucleotide signatures are the same, but the trinucleotide context provides far more useful information: for example, the spikes in AA-exposed UTUCs show that the AA-induced A > T mutations tend to occur in a (C|T)AG trinucleotide context. Reproduced from [76] with permission from Elsevier.