Machine Learning Meets Physics-based Modeling: A Mass-spring System to Predict Protein-ligand Binding Affinity.

Area of Science:

• Computational chemistry and structural biology.
• Drug discovery and development.

Background:

• Accurate computational assessment of protein-ligand binding energetics is crucial for early-stage drug discovery.
• Targeted scoring functions demonstrate superior performance over universal models in predicting binding affinity.

Purpose of the Study:

• To review the application of a simple physics-based mass-spring model for estimating binding affinity.
• To evaluate this model's predictive performance specifically for cyclin-dependent kinase inhibitors.

Main Methods:

• Literature search on PubMed for mass-spring models predicting binding affinity.
• Utilized crystal structures of cyclin-dependent kinases from the Protein Data Bank.
• Employed web servers for affinity calculations based on atomic coordinates.

Main Results:

• The Taba scoring function, a simple physics-based model, effectively analyzes protein-ligand interactions.
• Taba demonstrated superior performance compared to established physics-based models in AutoDock4 and Molegro Virtual Docker.
• Analysis of 27 scoring functions confirmed Taba's superior predictive metrics for cyclin-dependent kinase.

Conclusions:

• Machine learning advancements and accessible libraries facilitate the development of accurate protein-ligand interaction models.
• Integrating a mass-spring system with machine learning yields a targeted scoring function with enhanced predictive power for pKi estimation.