Abstract
Colorectal and gastric cancers are a leading cause of cancer deaths in developed countries. Precise estimation of prognosis is important with regard to clinical decision making for individuals with such cancers. We here comprehensively compiled a complete atlas of prognostic genes based on an integrated meta-analysis of one of the largest assembled colorectal cancer cohorts. A simple yet robust machine learning approach was then applied to establish a universal molecular prognostic score (mPS_colon) that relies on the expression status of only 16 genes and which was validated with independent data sets. This score was found to be an independent prognostic indicator in multivariate models including cancer stage, to be valid independent of tumor characteristics or patient ethnicity, and to be also applicable to gastric cancer. We conclude that mPS_colon is a universal prognostic classifier for patients with gastrointestinal cancers and that it should prove informative for optimization of personalized therapy for such patients.
Highlights
Cancer is the leading cause of death in developed countries, with an estimated 1.8 million new cases expected in the United States alone in 20201
Genes that are differentially expressed between normal mucosa and colorectal cancer (CRC) tissue (DEGs), and (2) genes that are associated with patient prognosis
We made use of public data to comprehensively identify prognosis-related genes with a meta-analysis of >1200 patients and we developed a universal prognostic classifier for gastrointestinal cancers, mPS_colon, with the use of machine learning technology
Summary
Cancer is the leading cause of death in developed countries, with an estimated 1.8 million new cases expected in the United States alone in 20201. Current TNM criteria, may give rise to substantial under- or overtreatment of individuals with CRC. Despite their receipt of similar treatment, CRC patients at the same stage show a wide range of outcomes. We hypothesized that such a difference in clinical outcome might be related to diverse transcriptome profiles of tumors. Identification of the molecular features of CRC that determine patient prognosis and stratification of patients on the basis of these features might be expected to inform the development of more effective clinical strategies and personalized therapies
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
