Computational advances in RNA-small molecule binding site prediction.

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology and Cheminformatics
• Drug Discovery and Medicinal Chemistry

Background:

• RNA-small molecule interactions are vital for cellular processes and represent promising therapeutic targets.
• Discovering RNA-binding molecules is difficult due to RNA's flexibility, dynamic binding sites, and lack of structural data.
• Computational approaches are essential for predicting these interactions.

Purpose of the Study:

• To review the evolution of computational strategies for predicting RNA-ligand binding sites.
• To highlight the integration of multimodal features and current challenges in the field.
• To discuss future directions for accurate and generalizable RNA-targeted drug discovery.

Main Methods:

• Evolution from statistical models to machine learning (ML) and deep learning (DL) frameworks.
• Integration of sequence, structural, energetic, and topological data.
• Application of large language models (LLMs) for sequence-based pattern recognition and multimodal modeling.

Main Results:

• ML/DL models now incorporate diverse data types for improved prediction accuracy.
• LLMs enhance the capture of long-range sequence dependencies and contextual information.
• Multimodal approaches combining sequence and structure information show promise.

Conclusions:

• Computational strategies have significantly advanced RNA-ligand binding site prediction.
• Integrating diverse data modalities, including LLMs, is key to overcoming existing challenges.
• Future work should focus on accuracy, generalizability, and interpretability to accelerate drug discovery.