Towards full Quantum-Mechanics-based Protein-Ligand Binding Affinities.

Area of Science:

• Computational chemistry
• Drug discovery
• Biophysics

Background:

• Current computational drug design relies on force fields with limited accuracy for complex interactions.
• Accurate prediction of protein-ligand binding affinities is crucial for pharmaceutical research.

Purpose of the Study:

• To develop and validate a general fully quantum mechanical (QM) scheme for computing protein-ligand binding free energies.
• To assess the QM method's performance on relevant pharmaceutical targets.

Main Methods:

• A QM scheme using a protein cutout (approx. 1000 atoms) to calculate absolute binding free energy.
• Energy minimization of model systems using graphics processing units (GPUs).
• High-level QM calculations, including hybrid DFT (PBEh-3c), for interaction, solvation, and thermostatistical contributions.

Main Results:

• The QM method achieved a mean absolute deviation of 2.1 kcal/mol for the FXa ligand set using the PBEh-3c composite method.
• GPU-accelerated energy minimization produced structures closely matching co-crystallized protein-ligand complexes.
• The approach successfully evaluated all contributions to absolute binding free energy.

Conclusions:

• The proposed QM scheme offers a significant improvement in accuracy for protein-ligand affinity calculations compared to force fields.
• Further improvements in structure optimization, conformational sampling, and solvation treatment are necessary for enhanced predictive power.
• This methodology holds potential for advancing computational drug design and discovery.