Absolute Binding Free Energies with OneOPES.

Area of Science:

• Computational chemistry
• Molecular dynamics simulations
• Drug discovery

Background:

• Calculating absolute binding free energies (ABFEs) for protein-ligand systems is crucial but historically challenging.
• Recent advancements in force fields and algorithms have improved ABFE calculation quality, yet accuracy for drug discovery remains a hurdle.

Purpose of the Study:

• To present a transferable enhanced sampling strategy for accurate ABFE calculations of protein-ligand systems.
• To validate the strategy using the OneOPES (One-particle Excitation Sampling) method with simple geometric collective variables.
• To assess the strategy's performance on diverse ligands and protein targets relevant to drug discovery.

Main Methods:

• Utilized the OneOPES enhanced sampling method combined with geometric collective variables.
• Applied the strategy to BRD4 and Hsp90 protein targets with 17 diverse ligands (fragments and drug-like molecules).
• Compared calculated binding free energies against experimental data and assessed sampling of binding modes and structural consistency.

Main Results:

• Achieved high accuracy in predicting protein-ligand binding affinities, with a mean unsigned error < 1 kcal mol⁻¹ compared to experimental values.
• Demonstrated the strategy's transferability across different protein targets and ligands without system-specific collective variable tuning.
• Showcased effective sampling of various ligand binding modes and consistent reproduction of experimentally determined structures from varied initial configurations.

Conclusions:

• The proposed OneOPES strategy offers an accurate and transferable approach for calculating protein-ligand ABFEs.
• This method can significantly inform lead optimization campaigns in drug discovery by providing reliable binding affinity predictions.
• The strategy facilitates deeper understanding of protein-ligand binding and unbinding mechanisms.