This study aimed to predict castration resistance in primary androgen deprivation therapy (ADT) for advanced prostate cancer using machine learning (ML) on integrated genetic and clinical data. Three ML algorithms (point-wise linear (PWL), logistic regression with elastic-net regularization, and XGBoost) were applied to data from the KYUCOG-1401-A study. The PWL algorithm, using clinical data and 46 SNPs, showed the highest area under the curve (AUC) for predicting castration resistance at 2 years. Models incorporating SNPs demonstrated superior prediction of castration resistance and prognosis compared to clinical data alone, suggesting their value in treatment decisions.
