A comparison of different initialization protocols to obtain statistically independent molecular dynamics simulations

Area of Science:

• Computational chemistry and molecular modeling.
• Biophysics and structural biology.

Background:

• Molecular dynamics (MD) simulations are crucial for understanding molecular interactions.
• Accurate setup of MD simulations is essential for reliable results, particularly in drug discovery.
• The MM/GBSA method is widely used for estimating ligand-protein binding affinities.

Purpose of the Study:

• To investigate the impact of various MD simulation setup choices on binding affinity calculations.
• To assess the reproducibility of MM/GBSA binding affinity predictions across different simulation parameters.
• To identify strategies for enhancing phase space sampling in MD simulations.

Main Methods:

• Calculated ligand-protein binding affinities using the MM/GBSA method for four protein targets.
• Varied MD simulation setup parameters including starting velocities, solvation, protonation states, and residue conformations.
• Analyzed the sensitivity of binding affinity results to these setup variations.

Main Results:

• MM/GBSA calculations demonstrated reasonable reproducibility across different MD simulation setups for some proteins (avidin, T4 lysozyme).
• Significant differences in binding affinity predictions (4-10 kJ/mol) were observed for other proteins (factor Xa, galectin-3) due to variations in solvation, conformations, or protonation states, especially near the binding site.
• Enhanced phase space sampling can be achieved by using diverse water boxes, starting velocities, and exploring alternative conformations and protonation states.

Conclusions:

• MM/GBSA binding affinity predictions are generally reproducible, but sensitive to setup choices for specific protein targets.
• Careful consideration of solvation, residue protonation, and conformational states is crucial for accurate predictions, particularly for residues near the active site.
• Employing multiple water boxes, starting velocities, and exploring available crystal conformations can enhance sampling and improve the reliability of MD-based binding affinity calculations.