RNA-targeted Drug Discovery: RNA-Small Molecule Binding Prediction

The drug discovery pipeline is a complex, multi-stage process encompassing target identification, hit discovery, lead optimization, and clinical development. While traditional drugs mainly target proteins, RNA-targeted drug discovery focuses on developing therapeutics that directly interact with RNA molecules. In this talk, we use the hit discovery stage as an example and introduce deep learning techniques designed for RNA–small molecule binding affinity and binding site prediction. DeepRSMA is a cross-attention-based deep learning model for predicting RNA–small molecule binding affinity. It employs nucleotide- and atomic-level feature extraction modules for RNA and small molecules, respectively, and uses a transformer-based cross-fusion module to model their interactions. The final prediction integrates features from both the extraction and fusion modules. MVRBind is a multi-view, multi-scale graph convolutional network for predicting RNA–small molecule binding sites. It captures RNA features at the primary, secondary, and tertiary structural levels and uses a fusion module to combine multi-scale representations into a unified embedding for accurate site prediction. Together, these models highlight the potential of deep learning to advance RNA-targeted drug discovery by improving the identification of promising RNA–small molecule interactions.