Development and validation of an individualized immune prognostic signature in colorectal cancer.

Abstract

e15531 Background: Colorectal cancer (CRC) is the third most common cause of cancer mortality worldwide with more than 1.85 million cases and 850,000 deaths annually. Although the outcome of CRC patients has improved significantly with the recent implementation of annual screening programs, reliable prognostic biomarkers are still needed due to the disease heterogeneity. Accumulating evidence revealed an association between immune signature and CRC prognosis. Therefore, we aim to build a robust immune-related gene pairs (IRGPs) signature that can evaluate the prognosis for CRC patients. Methods: Gene expression profiles and complete clinical information of CRC patients were collected from seven public cohorts, divided into the training cohort (TCGA cohort, N = 460) and six independent validation cohorts (GSE17536, GSE39582, GSE30378, GSE41258, GSE71187 and GSE87211, N = 1,151). Gene expression level in a specific sample or profile underwent pairwise comparison to generate a score for each IRGP. An IRGP score of 1 was assigned if IRG 1 was less than IRG 2; otherwise, the IRGP score was 0. Prognostic IRGPs were selected using log-rank test and LASSO Cox proportional hazards regression model. Kaplan-Meier curves and receiver operating characteristic (ROC) curves were used to examine model predictive performance in the validation cohort. Results: Within 942 immune-related genes that were expressed in all the cohorts, a prognostic immune signature of 80 gene pairs consisting of 101 unique genes was constructed which was significantly associated with patients overall survival in the training dataset (HR, 4.53 [2.89-7.09]; P < 0.0001). In the validation cohorts, the IRGPs signature significantly stratified patients into high-risk and low-risk groups in terms of prognosis and was prognostic in multivariate analyses (HR, 3.08 [2.03-4.65]; P < 0.0001). Biological processes that were dramatically involved in cancer immunology, such as cell chemotaxis and leukocyte migration, were enriched among genes in the IRGPs signature. To further improve accuracy, we combined tumor stage and IRGPs score to fit a Cox proportional hazards regression model using the training dataset and derived an immune-clinical prognostic index (ICPI) as (0.53633 × stage) + (1.31369 × IRGPs score). Significantly improved estimation of survival was achieved by ICPI relative to IRGPs score (mean C-index, 0.81 vs 0.79; P = 0 .014). Conclusions: We developed a robust IRGPs signature for estimating prognosis in CRC patients, providing new insights into the identification of CRC patients with a high risk of mortality. Prospective studies are needed to further validate its analytical accuracy and to test its clinical utility in individualized management of CRC patients.