Overfill Protection and Hyperdynamics in Adaptively Biased Simulations

Area of Science:

• Computational Chemistry
• Molecular Dynamics Simulations
• Statistical Mechanics

Background:

• Adaptive biasing methods in molecular simulations can suffer from overfilling and inaccurate time scale estimation.
• Controlling bias accumulation is crucial for efficient free energy calculations and enhanced sampling.

Purpose of the Study:

• To introduce a computationally efficient procedure for limiting bias fill-depth in adaptive biasing potentials.
• To enable accurate time scale estimation using hyperdynamics combined with depth-limited bias.
• To improve the reliability and efficiency of molecular dynamics simulations.

Main Methods:

• Development of a depth-limiting procedure for adaptive biasing potentials.
• Integration of depth-limited bias with hyperdynamics for time scale estimation.
• Testing on alanine dipeptide and ligand residence simulations.
• Evaluation of bias control strategies (fill-depth vs. update frequency).

Main Results:

• The introduced bias potential limits fill-depth, preventing accumulation in higher free energy regions.
• Depth-controlled bias yielded 6-8 times larger usable boost factors compared to update frequency control.
• The Kolmogorov-Smirnov test was found insufficient for hyperdynamics simulation validation; a robust strategy was described.

Conclusions:

• Directly controlling bias fill-depth is more effective than controlling update frequency for adaptive biasing and hyperdynamics.
• The developed methods enhance the efficiency and reliability of molecular dynamics simulations for complex systems.
• These advancements have been incorporated into the fABMACS project.