EISA-Score: Element Interactive Surface Area Score for Protein-Ligand Binding Affinity Prediction.

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Molecular surface representations are vital for studying protein structure and function, particularly in protein-ligand binding affinity modeling.
• Conventional surface-area-based methods often fall short in energy scoring due to insufficient encoding of physical and chemical interactions.
• There is a need for improved molecular surface representations that capture essential interactions for accurate binding affinity prediction.

Purpose of the Study:

• To develop novel molecular surface representations that accurately encode physical and chemical interactions.
• To introduce a new scoring function, the element interactive surface area-based scoring function (EISA-score), for protein-ligand binding affinity.
• To enhance the performance of machine learning algorithms in predicting structure-activity relationships.

Main Methods:

• Developed novel molecular surface representations using element interactive manifolds with dimensional reduction.
• Embedded accurate physical and biological property encoders within these low-dimensional descriptors.
• Paired these descriptors with machine learning algorithms to create the EISA-score.

Main Results:

• The novel representations achieved dramatic dimensional reduction while preserving crucial properties.
• The EISA-score demonstrated superior performance compared to state-of-the-art models.
• The EISA-score outperformed established surface-related representations on standard PDBbind benchmarks.

Conclusions:

• The proposed molecular surface representations effectively capture essential interactions for improved scoring.
• The EISA-score represents a significant advancement in protein-ligand binding affinity modeling.
• This approach offers a powerful tool for exploring structure-activity relationships in drug discovery.