Unbiased Enhanced Sampling in Molecular Dynamics via Simultaneously Accelerating Diffusion in Multiple...

Area of Science:

• Computational Chemistry
• Biophysics
• Statistical Mechanics

Background:

• Molecular dynamics (MD) simulations struggle with rare events due to high energy barriers.
• Enhanced sampling methods often require predefined collective variables (CVs), which are difficult to identify for complex systems.

Purpose of the Study:

• To develop a novel, unbiased enhanced sampling methodology for molecular dynamics (MD) simulations.
• To overcome the challenge of identifying optimal collective variables (CVs) for complex systems.

Main Methods:

• An iterative approach projecting unbiased sampling data onto multiple low-dimensional CV spaces.
• Calculating and integrating sampling density distributions to guide subsequent MD sampling.
• Simultaneously accelerating diffusive exploration in multiple relevant CV spaces.

Main Results:

• The method overcomes exponential efficiency decay in high-dimensional CV spaces.
• It eliminates the need for replica exchange by applying CV guidance to a single unbiased ensemble.
• Demonstrated accurate extraction of thermodynamic and kinetic properties for complex systems.

Conclusions:

• The proposed method offers a new paradigm for enhanced sampling in MD simulations.
• It synergistically combines unbiased sampling with CV algorithms for improved efficiency and accuracy.
• The approach facilitates on-the-fly CV generation and provides high-quality unbiased ensembles.