DeepRNA-DTI: a deep learning approach for RNA-compound interaction prediction with binding site interpretability
Haelee Bae, Hojung Nam · Journal of Cheminformatics · 2025
RNA-targeted therapeutics represent a promising frontier for expanding the druggable genome beyond conventional protein targets. However, computational prediction of RNA-compound interactions remains challenging due to limited experimental data and the inherent complexity of RNA structures. Here, we present DeepRNA-DTI, a novel sequence-based deep learning approach for RNA-compound interaction prediction with binding site interpretability. Our model leverages transfer learning from pretrained embeddings, RNA-FM for RNA sequences and Mole-BERT for compounds, and employs a multitask learning framework that simultaneously predicts both presence of interactions and nucleotide-level binding sites. This dual prediction strategy provides mechanistic insights into RNA-compound recognition patterns. Trained on a comprehensive dataset integrating resources from the Protein Data Bank and literature sources, DeepRNA-DTI demonstrates superior performance compared to existing methods. The model shows consistent effectiveness across diverse RNA subtypes, highlighting its robust generalization capabilities. Application to high-throughput virtual screening of over 48 million compounds against oncogenic pre-miR-21 successfully identified known binders and novel chemical scaffolds with RNA-specific physicochemical properties. By combining sequence-based predictions with binding site interpretability, DeepRNA-DTI advances our ability to identify promising RNA-targeting compounds and offers new opportunities for RNA-directed drug discovery. The codes and data are publicly available at https://github.com/GIST-CSBL/DeepRNA-DTI/. We present DeepRNA-DTI, a novel sequence-based deep learning framework that simultaneously predicts RNA-compound interactions and identifies nucleotide-level binding sites through multitask learning, providing mechanistic interpretability unlike existing methods that focus solely on interaction prediction. By integrating comprehensive datasets from PDB and literature sources with pretrained embeddings, our model demonstrates superior generalization across diverse RNA subtypes without requiring subtype-specific training. Large-scale virtual screening against oncogenic pre-miR-21 successfully identifies both known binders and novel RNA-specific chemical scaffolds, demonstrating practical utility for RNA-targeted drug discovery.