Randomized iterative trajectory reweighting for steady-state distributions without discretization error

Area of Science:

• Computational Chemistry
• Statistical Mechanics
• Molecular Dynamics Simulations

Background:

• Ensuring convergence to stationary distributions in molecular dynamics (MD) simulations is crucial for accurate estimation of thermodynamic properties and transition pathways.
• Unconverged simulations limit the reliability of calculated free energies, reaction rates, and mechanisms.

Purpose of the Study:

• To introduce a novel algorithm, Randomized Iterative trajectory Reweighting (RiteWeight), for estimating stationary distributions from unconverged MD simulation data.
• To address the challenge of configuration-space discretization errors in existing reweighting techniques.

Main Methods:

• The RiteWeight algorithm iteratively reweights trajectory segments by solving for the stationary distribution of a Markov state model (MSM).
• It employs self-consistent updates of segment weights and utilizes a novel random clustering approach in each iteration.
• The method mitigates discretization errors, yielding quasi-continuous configuration-space distributions without requiring the Markov property at the cluster level.

Main Results:

• RiteWeight accurately recovers the stationary distribution from both synthetic and atomistic MD trajectories.
• The algorithm successfully corrects flawed distributions obtained from unconverged simulations.
• Accurate observables for both equilibrium and nonequilibrium steady states were generated, demonstrating the method's efficacy.

Conclusions:

• RiteWeight provides a robust solution for estimating stationary distributions from unconverged MD data.
• The algorithm's ability to generate accurate observables highlights the importance of correcting the underlying trajectory distribution.
• This approach offers a valuable alternative to standard Markov State Models for analyzing molecular dynamics simulations.