Machine learning-based autophagy-related prognostic signature for personalized risk stratification and therapeutic approaches in bladder cancer.

Objective: Bladder cancer (BCa) is a highly lethal urological malignancy characterized by its notable histological heterogeneity. Autophagy has swiftly emerged as a diagnostic and prognostic biomarker in diverse cancer types. Nonetheless, the currently accessible autophagy-related signature specific to BCa remains limited.

Methods: A refined autophagy-related signature was developed through a 10-fold cross-validation framework, incorporating 101 combinations of machine learning algorithms. The performance of this signature in predicting prognosis and response to immunotherapy was thoroughly evaluated, along with an exploration of potential drug targets and compounds. In vitro and in vivo experiments were conducted to verify the regulatory mechanism of hub gene.

Results: The autophagy-related prognostic signature (ARPS) has exhibited superior performance in predicting the prognosis of BCa compared to the majority of clinical features and other developed markers. Higher ARPS is associated with poorer prognosis and reduced sensitivity to immunotherapy. Four potential targets and five therapeutic agents were screened for patients in the high-ARPS group. In vitro and vivo experiments have confirmed that FKBP9 promotes the proliferation, invasion, and metastasis of BCa.

Conclusions: Overall, our study developed a valuable tool to optimize risk stratification and decision-making for BCa patients.