LAUSR

OBJECTIVE Traditional diagnostic strategies are unable to accurately discriminate between patients with poor and satisfied prognosis in colon cancer. Therefore, it is urgently recommended to identify new biomarkers in favor of better selection of patients at higher risk of recurrence or poor outcomes, with the aim of early intervention or avoiding overtreatment. MATERIALS AND METHODS The weighted gene correlation network analysis (WGCNA), together with the proportion of tumor infiltrating immune cells, were employed to screen the key module related to immune infiltration. Using these genes among the key module, a predictive signature was generated via LASSO and multi-Cox regression method. Moreover, a novel nomogram was further developed by combining important clinical parameters and the predictive signature. RESULTS Genes among the green module, indicating the highest correlation with regulatory T cells (Tregs), were incorporated into the establishment of predictive model. Then, a Tregs-related risk signature (TRRS) consisting of four genes (NRG1, TEX11, OVOL3 and FCRL2) was established, which performed well in predicting the mortality risk of colon cancer in both internal and external validation groups (p=0.004 for TCGA training set, p=0.016 for TCGA testing set and p=0.03 for GSE39582 dataset). Combining TNM stage and age, we developed a nomogram for 1-, 3-, 5-year OS, presenting a more reliable predictive performance in survival based on the receiver operating characteristic (ROC) curves and calibration curves (3-year AUC: 0.83 and 0.74 in the TCGA and GEO database, respectively). CONCLUSIONS We constructed a four-gene signature for predicting the prognosis of patients with colon cancer, and further developed the nomogram together with TNM stage and age to improve the predictive efficacy.