Inference of non-exponential kinetics through stochastic resetting

Area of Science:

• Computational Chemistry
• Chemical Kinetics
• Molecular Dynamics

Background:

• Standard molecular dynamics simulations are limited to short timescales (∼1μs).
• Enhanced sampling methods and inference schemes are needed for slower processes.
• Existing inference schemes often assume exponential kinetics, unsuitable for non-exponential processes like power-law decay.

Purpose of the Study:

• To develop a novel inference scheme for long-timescale, non-exponential kinetics.
• To utilize molecular dynamics simulations accelerated by stochastic resetting.
• To overcome the limitations of current kinetics inference methods.

Main Methods:

• Simulations enhanced by stochastic resetting were employed.
• The scheme infers kinetics from the first-passage time distribution.
• It leverages short-timescale sampling and long-time asymptotic estimation.

Main Results:

• Stochastic resetting effectively samples the first-passage time distribution.
• The method successfully estimates long-time asymptotics for kinetics inference.
• Applied to a model system and a peptide, it achieved over an order of magnitude acceleration.
• Unbiased mean first-passage times were accurately estimated.

Conclusions:

• The proposed inference scheme accurately captures non-exponential kinetics.
• Stochastic resetting accelerates sampling and enables reliable kinetics estimation.
• This method advances the study of slow chemical processes using molecular dynamics.