Figure 1: Mutational load and mutational signatures in 49 gastric cancers.

(a) Mutational load: stacked bar plot of the single-nucleotide variant (SNV) count (y-axis) and nucleotide change with the changes indicated by the following colours: C>A (magenta), C>G (blue), C>T (green), T>A (yellow), T>C (orange) and T>G (grey). The tumour samples (x-axis) are ordered by the number of nucleotide variants, and each patient sample identifier is annotated with sex (M/F), EBV status (+: positive, −: negative), stage and Lauren histological classification (intestinal or diffuse). (b,c) Stacked bar plot of the SNV count and nucleotide changes in different trinucleotide sequence contexts for whole-genome (b) and -exome (c) regions. The flanking base pairs are arranged in the 5′ to 3′ direction in 16 categories (x-axis), for a total of 96 possible flanking nucleotide and SNV combinations. (d) Hierarchical cluster of 96 possible combinations of the flanking sequences and SNVs. The x-axis: tumour sample; y-axis: the trinucleotide context and corresponding SNVs. The SNV counts were converted to fractions to avoid co-clustering of samples with high single-nucleotide variant counts. The tumour samples largely cluster into two groups as marked by the yellow and blue bars. Three mutational patterns emerge: (1) TpT nucleotide, NpTpT>NpGpT and TpTpT>TpCpT (mostly impacting a subgroup of the diffuse tumour subtype, with a high mutational load); (2) CpG island-related regions, CpG>TpG and (3) TpCp[A/T]>TpTp[A/T] and TpCp[A/T]>TpAp[A/T], representing APOBEC mutational patterns related to TCW motifs (W=A/T). Samples with a CpG mutational have a reduced presence of the TpT-related pattern. The mutational load in the exonic, intronic and intergenic regions is provided in the lower panel, with samples arranged in the same order as the hierarchical cluster above. Samples with the TpT mutational pattern (highlighted by a blue bar) tended to occur largely in a subset of diffuse-type GCs and showed a much higher mutation rate in intergenic and intronic regions than in coding regions compared with the remaining samples.