As a cutting-edge technology, deep learning has been widely applied in biomedical research, bringing revolutionary changes to biology and medicine. By integrating genomic data with deep learning, it is possible to accurately identify high-risk groups for diseases and provide recommendations for subsequent diagnosis and treatment. Prostate cancer, one of the most common malignancies affecting older males, has long posed a challenge in terms of precisely identifying individuals at high risk. However, the performance of current risk assessment models remains limited. In this study, we employed a deep learning model, the Deep and Cross Network (DCN), to assess prostate cancer risk. This model is capable of capturing potential interactions within genetic features as well as between genetic and non-genetic factors. We developed two prostate cancer risk prediction models: one based solely on genetic data and another that integrates both genetic and non-genetic factors. In both cases, the DCN model demonstrated superior predictive performance compared to traditional methods. These results suggest that there are widespread interactions both among genetic features and between genetic and environmental factors, which may provide valuable insights into the underlying mechanisms of prostate cancer.