Molecular classification and risk stratification of colorectal cancer

IntroductionCurrently, there are no widely accepted molecular classification or prognostication model for colorectal cancer (CRC). To develop novel molecular classification and prognostication systems, we conducted large-scale studies of gene mutations involving 3056 CRC patients.MethodsAll patients were analyzed for mutations in 169 known/putative driver genes in CRC using targeted-capture sequencing, which also include 1,630 SNPs sequenced to assess genome-wide copy numbers.ResultsBased on the mutually exclusive and co-occuring mutation patterns of mutations in Wnt pathway, non-hypermutated CRC patients were classified into APC-mutated and unmutated case. The former was further divided into TP53-mutated and unmutated, while the latter was into CTNNB1-mutated, RNF43-mutated, RSPO-fusion-positive, and other cases. The APC/TP53-double mutated ‘canonical CRC’ cases were classified into 13 unique subtypes having unique mutational profiles using NMF clustering. Next we modeled time to recurrence (TTR) using Cox proportional hazard modeling. The final prediction model included gene mutations and copy number abnormalities, together with age, sex, stage, and tumor location, based on which we classified patients into groups having distinct TTR. Importantly, adjuvant chemotherapy (ACT) was shown to have significantly different impacts on TTR between different risk groups.ConclusionsOn the basis of comprehensive mutation analysis, we developed novel molecular classification of CRC. We further identified distinct risk groups showing different TTR, which could be used for optimized use of ACT.