Stochastic Lag Time Parameterization for Markov State Models of Protein Dynamics

Area of Science:

• Computational Biology
• Biophysics
• Statistical Mechanics

Background:

• Markov state models (MSMs) are crucial for understanding protein conformational dynamics.
• Traditional MSM construction relies on fixed lag times for sub-trajectory sampling, which can limit robustness and require extensive prior knowledge.
• Sub-optimal lag time selection can lead to inaccurate estimations of transition pathways and rates.

Purpose of the Study:

• To develop a more robust and less parameter-dependent method for constructing Markov state models.
• To improve the accuracy and reliability of MSMs, especially for systems exhibiting slow dynamics.
• To introduce a novel stochastic approach for sampling sub-trajectories in molecular dynamics.

Main Methods:

• A novel stochastic method based on a Poisson process is proposed to generate perturbative lag times for sub-trajectory sampling.
• This method is used to construct a Markov chain, forming the basis of the improved MSM.
• The algorithm's performance is evaluated on benchmark systems (double-well) and complex biological systems (WW domain, BPTI, RBD-ACE2 complex).

Main Results:

• The proposed method significantly enhances the robustness and statistical power of constructed MSMs.
• The algorithm maintains the essential Markovian properties of the system dynamics.
• Evaluations demonstrate superior performance compared to fixed lag time methods, particularly for slow dynamic modes.

Conclusions:

• The novel stochastic sampling method provides a more reliable approach to building Markov state models.
• This technique improves the analysis of protein conformational dynamics, especially in complex biological systems.
• The method offers increased robustness and accuracy without compromising the underlying Markovian assumptions.