Alchemical Free Energy Estimators and Molecular Dynamics Engines: Accuracy, Precision, and Reproducibility

Area of Science:

• Computational chemistry
• Molecular dynamics simulations
• Drug discovery

Background:

• Binding free energy is crucial for drug discovery.
• Computational assessment guides therapeutic development.
• Reproducibility of these calculations is key.

Purpose of the Study:

• To assess the reproducibility of alchemical relative binding free energy computations.
• To compare popular molecular dynamics packages and free energy estimators.
• To evaluate performance across diverse protein targets and ligand transformations.

Main Methods:

• Alchemical relative binding free energy calculations.
• Utilized OpenMM, NAMD2, and NAMD3 alpha molecular dynamics packages.
• Employed thermodynamic integration and free energy perturbation methods.

Main Results:

• Good agreement between molecular dynamics packages (average mean unsigned error of 0.50 kcal/mol).
• High correlation coefficients (Spearman's ≥ 0.92, Pearson's ≥ 0.91, Kendall's tau ≥ 0.76).
• Excellent agreement between thermodynamic integration and free energy perturbation with ensemble averaging.

Conclusions:

• Alchemical free energy calculations are reproducible across different software.
• This consistency supports the use of these methods in drug discovery pipelines.
• Ensemble averaging improves agreement between different free energy calculation methods.