Advances in computational prediction of RNA-small molecule binding affinity.

The burgeoning field of RNA-targeted drug discovery necessitates robust and accurate computational methods for predicting RNA-small molecule binding affinity. This review synthesizes recent advancements in deep learning and machine learning approaches, highlighting their methodologies, performance, and impact on accelerating drug design. We delve into methods that leverage diverse data representations, including sequence-based features, 3D structural information (voxel grids and molecular surfaces), and sophisticated graph-based networks. Key innovations such as contrastive learning, multi-scale feature extraction, and cross-fusion mechanisms are discussed, alongside their contributions to model robustness, generalization, and interpretability. We also consider the relative computational demands of these advanced models. Despite tremendous advancements, problems still exist, especially with regard to the lack of data because of the intrinsic flexibility of RNA structures and the inherent experimental difficulty in determining their structure and dynamic nature. The present state of computational RNA-small molecule affinity prediction is thoroughly reviewed in this article, along with important limits and future directions.