Autophagy-Related Three-Gene Prognostic Signature for Predicting Survival in Esophageal Squamous Cell Carcinoma

Abstract

Esophageal squamous cell carcinoma (ESCC) is one of the most aggressive malignant tumors in China, and its prognosis remains poor. Autophagy is an evolutionarily conserved catabolic process involved in the occurrence and development of ESCC. In this study, we described the expression profile of autophagy-related genes (ARGs) in ESCC and developed a prognostic prediction model for ESCC patients based on the expression pattern of ARGs. We used four ESCC cohorts, GSE53624 (119 samples) set as the discovery cohort, The Cancer Genome Atlas (TCGA) ESCC set (95 samples) as the validation cohort, 155 ESCC cohort, and Oncomine cohort were used to screen and verify differentially expressed ARGs. We identified 34 differentially expressed genes out of 222 ARGs. In the discovery cohort, we divided ESCC patients into three groups that showed significant differences in prognosis. Then, we analyzed the prognosis of 34 differentially expressed ARGs. Three genes [poly (ADP-ribose) polymerase 1 (PARP1), integrin alpha-6 (ITGA6), and Fas-associated death domain (FADD)] were ultimately obtained through random forest feature selection and were constructed as an ARG-related prognostic model. This model was further validated in TCGA ESCC set. Cox regression analysis confirmed that the three-gene signature was an independent prognostic factor for ESCC patients. This signature effectively stratified patients in both discovery and validation cohorts by overall survival (P = 5.162E-8 and P = 0.052, respectively). We also constructed a clinical nomogram with a concordance index of 0.713 to predict the survival possibility of ESCC patients by integrating clinical characteristics and the ARG signature. The calibration curves substantiated fine concordance between nomogram prediction and actual observation. In conclusion, we constructed a new ARG-related prognostic model, which shows the potential to improve the ability of individualized prognosis prediction in ESCC.