Prediction of ligand binding mode among multiple cross-docking poses by molecular dynamics simulations

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Accurately predicting protein-ligand binding modes is crucial for rational drug design.
• Distinguishing correct poses from numerous false positives generated by docking is a significant challenge.

Purpose of the Study:

• To develop and validate a robust computational method for identifying the correct ligand binding pose among multiple docking predictions.
• To improve the accuracy of molecular docking by integrating molecular dynamics (MD) simulations and binding free energy calculations.

Main Methods:

• Utilizing five independent MD simulations with varied initial velocities to assess the stability of each docking pose.
• Filtering out unstable poses based on root-mean-square deviation (RMSD) thresholds (2.0 Å).
• Performing all-atom binding free energy calculations on the remaining stable poses to determine the most accurate binding mode.

Main Results:

• The proposed method successfully identified the correct ligand binding mode for 72% (75 out of 104) of the complexes in a cross-docking test set.
• The method effectively filters out incorrect poses using MD simulations before computationally intensive free energy calculations.

Conclusions:

• The combined approach of MD simulations and binding free energy calculations offers a reliable strategy for validating protein-ligand docking poses.
• Further investigation into failure cases is necessary to refine the method and enhance its accuracy for broader applications in drug discovery.