Predicting absolute ligand binding free energies to a simple model site

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Accurate prediction of binding free energies is a key challenge in structure-based ligand design.
• Alchemical free energy calculations offer a promising approach for this prediction.

Purpose of the Study:

• To apply alchemical free energy calculations in explicit solvent to predict ligand binding in a model cavity of T4 lysozyme.
• To systematically improve prediction accuracy by incorporating multiple poses and protein conformational changes.
• To evaluate the performance of these methods in blind prospective tests.

Main Methods:

• Alchemical free energy calculations in explicit solvent.
• Inclusion of multiple ligand poses from docking.
• Accounting for protein conformational flexibility.
• Use of an improved protein charge model.

Main Results:

• Computed absolute binding free energies showed an RMS error of 1.9 kcal/mol against experimental values.
• Prospective tests demonstrated correct discrimination between ligands and decoys.
• Prospective RMS error in binding free energy prediction was 0.6 kcal/mol.
• X-ray structures validated free energy calculation predictions, sometimes differing from docking predictions.

Conclusions:

• Alchemical free energy calculations, with systematic improvements, can accurately predict ligand binding free energies.
• These methods show strong performance in prospective validation, outperforming docking in some structural aspects.
• Understanding the impact of protein rigidity approximations is crucial for improving docking accuracy.