Computational methods for binding site prediction on macromolecules

Area of Science:

• Biochemistry and Structural Biology
• Computational Chemistry
• Pharmacology

Background:

• Binding sites on biomolecules (proteins, RNAs) are crucial for molecular interactions.
• Experimental identification of binding sites is costly and slow.
• Computational methods offer efficient alternatives for binding site analysis.

Purpose of the Study:

• To review recent advancements in computational binding site identification.
• To classify machine learning approaches based on data encoding and interacting molecule types.
• To discuss future perspectives and challenges in the field.

Main Methods:

• Classification of machine learning methods based on macromolecule information (sequence, structure, geometry, energy).
• Categorization of methods by the type of interacting molecule (small molecules, peptides, ions).
• Emphasis on deep learning-based approaches for binding site prediction.

Main Results:

• Recent machine learning methods provide scalable and efficient binding site identification.
• Diverse computational strategies exist, leveraging various data types for prediction.
• Deep learning shows promise for advancing binding site analysis.

Conclusions:

• Computational methods, particularly machine learning, are vital for modern drug discovery.
• Identifying novel binding sites expands the potential for therapeutic interventions.
• Further development of deep learning models is expected to enhance drug design and optimization.