RBSP-CAN: RNA Binding Site Prediction by Convolution and Attention Network

RNA molecules play a crucial role in regulating and catalyzing biological processes and are closely linked to the development of numerous diseases, including neurological disorders and cancer. To achieve their biological regulatory functions, most RNA molecules require binding to other small molecules. Consequently, predicting the binding sites of RNA and small molecules is essential for the research of targeted drug development for RNA. However, only a limited number of relevant methods have been proposed thus far, and predicting RNA-small molecule binding sites remains a challenging task. To improve our ability to predict such binding sites, we require better models that can integrate RNA features more effectively. Those current computational models do not fully leverage the sequence features of RNA. In this paper, we propose a deep learning model, RBSP-CAN, to effectively predict RNA-small molecule binding sites by utilizing attention and convolution mechanisms that focus on the sequence features of RNA. The experimental results demonstrate that RBSP-CAN outperforms other state-of-the-art methods in predicting binding sites.