Identification of de novo signatures from human colorectal and gastric adenocarcinoma samples (COAD-US and STAD-US projects) and their contribution to samples clinically classified as microsatellite instable (MSI) or microsatellite stable (MSS). (A) Two-dimensional representation of the mutational profile composition across cancer samples. The size of each circle reflects the mutation burden. MSI samples are highlighted by a bold, black outline. The color of segments reflects the signature composition. (B) Mutational signatures including base substitutions and 1-bp indels derived from the combined COAD-US and STAD-US data sets. (C) Relative contribution of MMR-1, MMR-2, and MMR-3 signatures to cancer samples clinically classified as MSI or MSS. Box plot with outliers shown as individual filled circles. Area under the curve (AUC) value for MMR-1 contribution indicates the probability of a random MSI sample having higher MMR-1 contribution than a random MSS sample. (D) Number of mutations assigned to signatures MMR-1 (green), MMR-2 (purple), and MMR-3 (orange) plotted against the number of 1-bp indels in the same sample. (E) Fold change in the average number of mutations assigned to different signatures in MSI samples compared to MSS samples. As expected, the number of POLE-related mutations is higher in MSS samples as all POLE-deficient tumors are MSS. Apart from MMR signatures, Clock-1 signature also contributes over 10 times more mutations to MSI samples than to MSS. SNP associated mutations are likely due to unfiltered SNPs that are prevalent in the human population (Supplemental Material).
