LAUSR

OBJECTIVE Cell metabolism plays a vital role in the proliferation, metastasis and sensitivity to chemotherapy drugs of colorectal cancer. The purpose of this multicenter cohort study is to investigate the potential genes indicating clinical outcomes in colorectal cancer patients. METHODS We analyzed gene expression profiles of colorectal cancer to identify differentially expressed genes then used these differentially expressed genes to construct prognostic signature based on the least absolute shrink-age and selection operator Cox regression model. In addition, the multi-gene signature was validated in independent datasets including our multicenter cohort. Finally, nomograms were set up to evaluate the prognosis of colorectal cancer patients. RESULTS Seventeen metabolism-related genes were determined in the least absolute shrink-age and selection operator model to construct signature, with area under receiver operating characteristic curve for relapse-free survival, 0.741, 0.755 and 0.732 at 1, 3 and 5 year, respectively. External validation datasets, GSE14333, GSE37892, GSE17538 and the Cancer Genome Atlas cohorts, were analyzed and stratified, indicating that the metabolism-related signature was reliable in discriminating high- and low-risk colorectal cancer patients. Area under receiver operating characteristic curves for relapse-free survival in our multicenter validation cohort were 0.801, 0.819 and 0.857 at 1, 3 and 5 year, respectively. Nomograms incorporating the genetic biomarkers and clinical pathological features were set up, which yielded good discrimination and calibration in the prediction of prognosis for colorectal cancer patients. CONCLUSION An original metabolism-related signature was developed as a predictive model for the prognosis of colorectal cancer patients. A nomogram based on the signature was advantageous to facilitate personalized counselling and treatment of colorectal cancer patients.