Ab initio sampling of transition paths by conditioned Langevin dynamics

Area of Science:

• Statistical physics
• Computational chemistry
• Biophysics

Background:

• Simulating molecular dynamics requires accurate path generation.
• Brownian paths are crucial for understanding transitions in complex potentials.
• Existing methods can be computationally intensive.

Purpose of the Study:

• To introduce a novel stochastic method for generating conditioned Brownian paths.
• To enable efficient and statistically independent sampling of transition paths.
• To apply the method to benchmark potentials and protein conformational changes.

Main Methods:

• Developing a stochastic method based on overdamped Langevin dynamics.
• Utilizing a local stochastic partial differential equation for path generation.
• Implementing approximations for low-temperature and barrier-crossing regimes.

Main Results:

• Demonstrated exact generation of conditioned Brownian paths.
• Achieved computationally efficient and statistically independent path generation.
• Successfully applied the method to Mueller, Mexican hat, and protein transition problems.

Conclusions:

• The proposed method offers an efficient way to simulate transition paths.
• It provides a robust tool for studying molecular dynamics and conformational changes.
• The technique is applicable to various complex multi-dimensional systems.