Identification of a 9-gene autophagy-related signature for predicting prognosis and immune exhaustion features in breast cancer

Abstract To address the prognostic limitations in breast cancer (BRCA), we integrated transcriptomic profiles from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) to construct a novel 9-gene autophagy-related signature via Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression. The model demonstrated robust predictive accuracy for overall survival in the training cohort (HR = 2.28, P < 0.001) and maintained stability in external validation (AUC = 0.740). Mechanistically, the risk score was significantly associated with selective autophagy pathways and an immune-exhausted microenvironment characterized by T-cell dysfunction. Furthermore, drug sensitivity profiling identified a positive correlation between the risk gene MTDH and sensitivity to Vincristine and Gemcitabine. This study presents a reliable risk-stratification tool that bridges autophagic mechanisms with personalized chemotherapy guidance.