Colorectal Cancer Subtyping Consortium (CRCSC) Identifies Consensus of Molecular Subtypes

Abstract

Background: Recently, a number of independent groups reported novel molecular subtypes in colorectal cancer (CRC). A formal comparison across these classifiers is needed to reconcile findings and accelerate clinical translation. The CRCSC was formed to identify a consensus among the subtyping systems through large scale data sharing and meta-analysis. Methods: The CRCSC consists of 6 groups (15+ institutions) that analyzed more than 30 patient cohorts with gene expression data, spanning multiple platforms and sample preparation methods. Each of the 6 classifiers (with 3-6 subtypes) was applied to the collection of public and proprietary datasets encompassing over 4,000 samples, mostly stage II-III CRC. Concordance of subtype calls and associations with clinical, molecular and pathway features were assessed centrally by an independent team. Results: Despite heterogeneities in cohorts and methods, subtype concordance analysis readily yielded a clear consensus on 4 CRC molecular subtypes (CMS1-4), with significant interconnectivity among the calls from the participating groups. The remaining 16% of samples did not have a consensus assignment, which may be partly explained by an additional mixed subtype with variable epithelial-mesenchymal activation; further refinement is needed. Conclusions: This is the first example of a large-scale, community based comparison of cancer subtypes. Within the largest collection of CRC samples we identified recurrent signals of 4 biologically distinct subtype classes enriched for key clinical, pathway and molecular traits. Ongoing efforts are attempting to improve the granularity of these subtypes.