Markov-Type State Models to Describe Non-Markovian Dynamics.

Area of Science:

• Computational Chemistry
• Biophysics
• Statistical Mechanics

Background:

• Molecular dynamics (MD) simulations are crucial for studying molecular behavior.
• Clustering MD trajectories into metastable states often violates the assumption of time scale separation.
• This violation complicates the construction of accurate Markov state models (MSMs).

Purpose of the Study:

• To address the challenge of inaccurate transition matrix estimation in MSMs when time scale separation is invalid.
• To evaluate advanced methods for constructing more reliable MSMs from MD data.
• To compare the performance of different approaches using toy models and real biological systems.

Main Methods:

• Laplace-transform-based method (Hummer and Szabo).
• Direct microstate-to-macrostate projection.
• Quasi-Markov state model (MSM) ansatz (Huang et al.).
• Hybrid method combining MD and MSM.

Main Results:

• Naive MSM construction leads to inaccurate time scales and population decays when time scale separation is absent.
• The evaluated advanced methods offer improved accuracy in estimating macrostate transition matrices.
• Each method demonstrates specific strengths and weaknesses when applied to different systems.

Conclusions:

• Accurate estimation of transition matrices is critical for reliable MSMs in MD simulations.
• Advanced methods provide viable solutions to overcome limitations imposed by violated time scale separation assumptions.
• The choice of method depends on the specific characteristics of the molecular system and simulation data.