Convolutional networks with parallel structure for metastatic prostate cancer prediction

Accurately predicting the future cancer status of prostate cancer patients is critical for treatment. Studies show a strong link between prostate cancer and genetic mutations. To better predict a patient's cancer status and identify key mutated genes during metastasis, we propose a convolutional network with parallel structure (CNPS). Our approach includes a mutation data preprocessing method for easier feature extraction, followed by parallel convolutional networks to capture gene mutation features across multiple dimensions for more accurate predictions. Finally, CNPS is highly interpretable, allowing us to identify key genes involved in metastatic prostate cancer. After training, CNPS achieves higher accuracy on both the MPC and MSK-MET datasets.