Current State of Open Source Force Fields in Protein-Ligand Binding Affinity Predictions.

Area of Science:

• Computational chemistry
• Drug discovery
• Molecular modeling

Background:

• In silico prediction of binding affinity is vital for prioritizing drug candidates.
• Alchemical relative binding free energy (RBFE) calculations using molecular dynamics (MD) simulations are a popular method for accurate affinity ranking.

Purpose of the Study:

• To evaluate the performance of six different small-molecule force fields in predicting experimental protein-ligand binding affinities.
• To assess the impact of force field parameters on prediction accuracy.

Main Methods:

• RBFE calculations using MD simulations were performed.
• Six small-molecule force fields (OpenFF Parsley, Sage, GAFF, CGenFF, OPLS3e) were evaluated.
• A dataset of 598 ligands and 22 protein targets was used.

Main Results:

• OPLS3e demonstrated significantly higher accuracy compared to other tested force fields.
• A consensus approach combining Sage, GAFF, and CGenFF achieved accuracy comparable to OPLS3e.
• Force field parameter improvements led to better accuracy for specific subsets.
• Input preparation and simulation convergence also impacted prediction accuracy.

Conclusions:

• Force field choice significantly impacts the accuracy of binding affinity predictions.
• A consensus strategy can enhance predictive power.
• Addressing input preparation and sampling convergence is essential for reliable in silico predictions.