Enhanced modeling via network theory: Adaptive sampling of Markov state models

Area of Science:

• Computational chemistry
• Biophysics
• Molecular dynamics

Background:

• Computer simulations offer atomic-level insights into molecular kinetics.
• Current supercomputers face limitations in simulating large biological systems over relevant timescales.

Purpose of the Study:

• To demonstrate the efficiency of adaptive sampling with Markov State Models (MSMs) for molecular simulations.
• To introduce a novel distance metric for comparing MSMs.

Main Methods:

• Utilized adaptive sampling techniques to build Markov State Models (MSMs).
• Developed and applied a relative entropy-based distance metric for MSM comparison.
• Assessed the convergence of sampling schemes using the new metric.

Main Results:

• Adaptive sampling with MSMs significantly reduces computational time and resources.
• Previously intractable molecular simulations become routinely feasible on existing hardware.
• The relative entropy metric effectively evaluates MSM convergence and system dynamics.

Conclusions:

• MSMs combined with adaptive sampling offer a powerful solution for computationally intensive molecular simulations.
• The developed distance metric is valuable for assessing simulation quality and system perturbations.
• This methodology advances the feasibility of studying complex biological systems at the molecular level.