Subtype-specific risk models for accurately predicting the prognosis of breast cancer using differentially expressed autophagy-related genes

Abstract

Emerging evidence suggests that the dysregulation of autophagy-related genes (ARGs) is coupled with the carcinogenesis and progression of breast cancer (BRCA). We constructed three subtype-specific risk models using differentially expressed ARGs. In Luminal, Her-2, and Basal-like BRCA, four-(BIRC5, PARP1, ATG9B, and TP63), three-(ITPR1, CCL2, and GAPDH), and five-gene (PRKN, FOS, BAX, IFNG, and EIF4EBP1) risk models were identified, which all have a receiver operating characteristic > 0.65 in the training and testing dataset. Multivariable Cox analysis showed that those risk models can accurately and independently predict the overall survival of BRCA patients. Comprehensive analysis showed that the 12 identified ARGs were correlated with the overall survival of BRCA patients; six of the ARGs (PARP1, TP63, CCL2, GAPDH, FOS, and EIF4EBP1) were differentially expressed between BRCA and normal breast tissue at the protein level. In addition, the 12 identified ARGs were highly interconnected and displayed high frequency of copy number variation in BRCA samples. Gene set enrichment analysis suggested that the deactivation of the immune system was the important driving force for the progression of Basal-like BRCA. This study demonstrated that the 12 ARG signatures were potential multi-dimensional biomarkers for the diagnosis, prognosis, and treatment of BRCA.