DRLiPS: a novel method for prediction of druggable RNA-small molecule binding pockets using machine learning

Area of Science:

• Computational Biology
• Drug Discovery
• Molecular Biology

Background:

• Ribonucleic Acid (RNA) plays a crucial role in cellular information transfer.
• Identifying druggable RNA targets for disease treatment is challenging due to the complexity of non-coding RNAs.
• Existing protein-druggability prediction methods are not suitable for RNA due to different binding mechanisms.

Purpose of the Study:

• To develop a structure-based model for predicting the druggability of RNA binding sites.
• To address the limitations of current methods in identifying RNA-centric therapeutic targets.
• To provide a computational tool for advancing RNA-targeted drug discovery.

Main Methods:

• Development of the Druggable RNA-Ligand binding Pocket Selector (DRLiPS) model.
• Utilized a novel strategy for sampling negative binding sites using backbone motif search, exhaustive pocket prediction, and blind docking.
• Curated an external blind test dataset comprising experimental and modeled apo state RNA structures.

Main Results:

• DRLiPS achieved an F1-score of 0.70, precision of 0.61, specificity of 0.89, and recall of 0.73 on the external test dataset.
• The model outperformed existing methods like DrugPred_RNA and RNACavityMiner.
• Selected features generalize well to both apo and holo RNA states and can predict the impact of mutations on druggability.

Conclusions:

• DRLiPS is an effective structure-based model for predicting RNA binding site druggability.
• The model demonstrates generalizability and outperforms existing tools, aiding in the identification of novel RNA drug targets.
• DRLiPS can assist in optimizing RNA aptamers for small molecule recognition and offers a freely accessible platform for researchers.