Computational analysis of mRNA expression profiles identifies a novel triple-biomarker model as prognostic predictor of stage II and III colorectal adenocarcinoma patients

Introduction: Although remarkable progress has been made to determine the prognosis of patients with colorectal cancer (CRC), it is inadequate to identify the subset of high-risk TNM stage II and stage III patients that have a high potential of developing tumor recurrence and may experience death. In this study, we aimed to develop biomarkers as a prognostic signature for the clinical outcome of CRC patients with stage II and stage III. Materials and methods: We performed a systematic and comprehensive discovery step to identify recurrence-associated genes in CRC patients through publicly available GSE41258 (n=253) and GSE17536 (n=107) datasets. We subsequently determined the prognostic relevance of candidate genes in stage II and III patients and developed a triple-biomarker for predicting RFS in GSE17536, which was later validated in an independent cohort GSE33113 dataset (n=90). Results: Based upon mRNA expression profiling studies, we identified 45 genes which differentially expressed in recurrent vs non-recurrent CRC patients. By using Cox proportional hazard models, we then developed a triple-marker model (THBS2, SERPINE1, and FN1) to predict prognosis in GSE17536, which successfully identified poor prognosis in stage II and stage III, particularly high-risk stage II CRC patients. Discussion: Notably, we found that our triple-marker model once again predicted recurrence in stage II patients in GSE33113. Kaplan-Meier survival analysis demonstrated that patients with high scores have a poor outcome compared to those with low scores. Our triple-marker model is a reliable predictive tool for determining prognosis in CRC patients with stage II and stage III, and might be able to identify high-risk patients that are candidates for more targeted personalized clinical management and surveillance.