VAD-MM-GBSA

VAD-MM-GBSA computes protein–ligand binding free energies using a variable atomic dielectric MM/GBSA framework to improve the accuracy of binding affinity predictions for structure-based drug design.

Key Features:

• Variable Dielectric Constants: Assigns variable dielectric constants directly to individual atoms in protein and ligand molecules to provide a more nuanced representation of electrostatic interactions within the MM/GBSA formalism.
• Machine Learning Optimization: Employs machine learning algorithms to optimize the assignment of atomic dielectric constants based on protein–ligand interaction data.
• Improved Predictive Performance: Demonstrates superior binding free energy prediction compared to traditional MM/GBSA implementations, including those in Schrödinger, with notable performance on challenging targets.
• Minimal Computational Overhead: Introduces only a slight increase in computational demand relative to standard MM/GBSA methods.

Scientific Applications:

• Structure-based drug design (SBDD): Enhances binding free energy estimates to support lead identification and optimization in SBDD workflows.
• Virtual screening postprocessing: Refines postprocessing stages of structure-based virtual screening to improve selection of viable candidate compounds.
• Target evaluation: Applied to challenging targets, including POL polyprotein and human immunodeficiency virus type 1 (HIV-1) protease, to assess predictive performance.
• Molecular interaction analysis: Provides atom-level electrostatic characterization useful for analyzing protein–ligand interactions.

Methodology:

Per-atom dielectric assignment within an MM/GBSA framework combined with simulated annealing for dielectric optimization, guided by machine learning algorithms trained on protein–ligand interaction data.