Abstract 978: Genomic characterization and risk stratification of colorectal cancer

Abstract Introduction: Adjuvant chemotherapy (ACT) has a key role in the treatment of high-risk colorectal cancer (CRC) patients to prevent post surgical recurrence. However, the efficacy of ACT is offset by the development of complication such as peripheral neuropathy, which leads to compromised quality of life, while ACT is expected to prevent tumor recurrence in only 5% and 20% in Stage2 and 3 patients, respectively. Thus, there is an urgent need to develop a biomarker that accurately predicts the effect of ACT and therefore helps guide ACT eligibility. Methods: We enrolled a total of 3,023 cases with CRC, for whom targeted-capture sequencing was performed that covered 169 known or putative driver genes and 1,660 polymorphoic sites to detect somatic mutations and copy number alterations (CNAs), respectively. RNA sequencing was also performed to detect driver gene fusions and to profile gene expression. We identified, through univariate analysis, gene mutations/CNAs as well as clinical parameters using those cases who did not received ACT to establish a model that predict time to recurrence (TTR) using the Cox proportional hazard modelling. Thereafter, the model was applied to those who did receive ACT to assess the impact of the model in the prediction of TTR. Results: We identified a total of 32145 mutations and/or 44511 CNAs in 2970 (98%) out of 3,023 cases. Mutations were found in 55 driver genes, including 15 previously unknown drivers, such as MAP2K1, AKT1, and TCF7. According to the number of mutations, 278 casese were classified as hypermutated cases. The remaining ‘non-hypermutated cases’ (n=2692) were classified into APC-mutated, RNF43/ZNRF3-mutated, CTNNB1-mutated, RSPO-fusion(+) and other Wnt-mutation(−) cases. The TTR prediction model was constructed by repeating Cox regression 1000 times using randomly selected 80% of the non-hypermutated cases and variables identified in univariate analysis. Adopting those variables selected in >70% models, we established the final model including both genetic (mutations/CNAs) and clinical variables, based on which we classified patients into 4 risk groups, low/intermediate/high/very-high. Finally, by applying the classification to those who received ACT, we demonstrated that 18% of Stage2 and the majority of Stage3 cases with high/very-high risks were predicted to benefit from ACT, whereas most of the Stage2 and as many as 19% of Stage3 cases received ACT with no significant improvement of TTR. Approximately 5% of Stage1 cases belonged to the high-risk category and potentially benefit from ACT. Conclusions: On the basis of large-scale genotyping of clinically well-annotated CRC cases, we proposed molecular classification of CRC and established a model that enabled prediction of TTR and risk stratification, which was successfully used to identify cases who benefited (within Stage 2/3 cases) or will benefit (in Stage1 cases) from ACT to prolong TTR. Citation Format: Yoshikage Inoue, Nobuyuki Kakiuchi, Kenichi Yoshida, Yasuhito Nanya, Yasuhide Takeuchi, Yuichi Shiraishi, Kenichi Chiba, Tetusichi Yoshizato, Hiroko Tanaka, Satoshi Nagayama, Kazutaka Obama, Satoru Miyano, Seishi Ogawa. Genomic characterization and risk stratification of colorectal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 978.