Efficient Bayesian estimation of Markov model transition matrices with given stationary distribution.

Area of Science:

• Computational Chemistry
• Biophysics
• Statistical Mechanics

Background:

• Direct simulation of biomolecular dynamics is computationally expensive and faces challenges with metastable conformational changes.
• Enhanced sampling methods can improve convergence for equilibrium probabilities and stationary quantities.
• Estimating dynamic observables like correlation functions requires direct equilibrium simulations, which are often slow to converge.

Purpose of the Study:

• To develop statistical estimation methods for incorporating a priori knowledge of equilibrium probabilities into the estimation of dynamical observables.
• To improve the convergence of dynamic observables in molecular dynamics simulations.
• To provide more efficient methods for analyzing slow dynamical processes in molecular systems.

Main Methods:

• Introduction of statistical estimation methods that leverage known equilibrium probabilities.
• Development of maximum likelihood methods for estimating transition matrices.
• Implementation of an improved Monte Carlo sampling method for reversible transition matrices with fixed stationary distributions.

Main Results:

• The proposed methods allow for the incorporation of prior equilibrium probability knowledge into dynamical observable estimation.
• Both maximum likelihood and improved Monte Carlo sampling methods are presented.
• The sampling approach demonstrated significantly faster convergence compared to previous methods when applied to a peptide system.

Conclusions:

• The developed statistical estimation methods enhance the analysis of biomolecular dynamics by improving convergence rates.
• These methods are well-suited for Markov state models, enabling better characterization of both stationary and dynamic properties.
• The approach offers a more efficient way to study slow conformational transitions and dynamic observables in molecular systems.