Abstract
BACKGROUND: Bladder cancer (BLCA) is an aggressive malignancy characterized by high rate of recurrence. Its steadily increasing incidence and prevalence have made it one of the most prevalent urogenital tract tumors worldwide. Although immunotherapy serves as a first-line treatment for BLCA, patient prognosis shows significant heterogeneity. Risk stratification through immunotherapy risk scoring could substantially improve prognostic predictions. However, current studies examining the use of immune treatment risk scores to predict the prognosis of immunotherapy across various patients with BLCA remain in its early stages. This research aims to identify immune-related genes as potential biomarkers for BLCA immunotherapy guidelines. METHODS: To fill this gap, we used two independent immunotherapy datasets from the "IMvigor210CoreBiologies" package as the training cohort and the Gene Expression Omnibus (GEO) database as the validation cohort to create a risk score signature based on deep learning algorithm to assess the effectiveness and outlook of BLCA immunotherapy. RESULTS: A risk score model comprising three immunotherapeutic-related genes was established and validated, demonstrating significant predictive power and serving as an independent factor for forecasting overall survival (OS) in BLCA immunotherapy. Furthermore, our model revealed a strong association with drug sensitivity responses and identified immune landscape differences among various BLCA patients. CONCLUSIONS: We anticipated that the risk score, which is an independent prognostic factor, would be taken into consideration when deciding whether to provide clinical immunotherapy for BLCA patients.