Identification and Validation of a Robust Autophagy-Related Molecular Model for Predicting the Prognosis of Breast Cancer Patients

Abstract

Despite a relatively low mortality rate, high recurrence rates represent a significant problem for breast cancer (BC) patients. Autophagy affects the development, progression, and prognosis of various cancers, including BC. The aim of the present study was to identify candidate autophagy-related genes (ARGs) and construct a molecular-clinicopathological signature to predict recurrence risk in BC. A 10-ARG-based signature was established in a training cohort (GEO-BC dataset GSE25066) with LASSO Cox regression and assessed in an independent validation cohort (GEO-BC GSE22219). Significant differences in recurrence-free survival were observed for high- and low-risk patients segregated based on their signature-based risk score. Time-dependent receiver operating characteristic (tdROC) analysis of signature performance demonstrated satisfactory accuracy and predictive power in both the training and validation cohorts. Moreover, we developed a nomogram to predict 3- and 5-year recurrence-free survival by combining the autophagy-related risk score and clinicopathological data. Both the tdROC and calibration curves indicated high discriminating ability for the nomogram. This study indicates that our ARG-based signature is an independent prognostic classifier for recurrence-free survival in BC. In addition, individualized survival risk assessment and treatment decisions might be effectively improved by implementing the proposed nomogram.