Abstract 103: Gene expression based subtypes of colorectal cancer

Abstract

Abstract Colorectal cancer is a heterogeneous disease and improved molecular stratification of patients is warranted for precise treatment. Several gene expression classification systems are reported in the literature, albeit many have shown low reproducibility in external validation. A robust random forest classification system based on gene expression profiles was recently proposed, suggesting the existence of four biologically different subtypes with associations to prognosis (CMS1, CMS2, CMS3 and CMS4) (Guinney,J., et al., Nat Med, 2015). By applying the CMS classifier to high-quality gene expression profiles from a consecutive series of >400 stage I-IV primary colorectal cancers we are able to reproduce results from Guinney et al in an independent cohort. Analyses of mutational hotspots in BRAF / KRAS and the complete coding sequence of TP53 confirms enrichment of BRAF mutations in CMS1 (p<0.001), TP53 mutations in CMS2 and KRAS mutations in CMS4 (p<0.001). We further validate high levels of somatic copy number aberrations (SCNAs) in CMS2 and CMS4 compared to CMS1 and CMS3 (pair-wise comparisons, fraction of genome affected by SCNAs: CMS1 vs CMS2: p<0.001; CMS1 vs CMS4: p<0.01; CMS2 vs CMS3: p<0.001; CMS3 vs CMS4: p<0.01). Additionally we explore within-subtype variations with respect to SCNAs in the CMS2 and CMS4 subtypes. Altogether our results validate the CMS system as a robust classification system for colorectal cancer. Citation Format: Kaja G. Berg, Ina A. Eilertsen, Sharmini Alagaratnam, Stine A. Danielsen, Arild Nesbakken, Anita Sveen, Ragnhild A. Lothe. Gene expression based subtypes of colorectal cancer. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 103.