Markov state models from hierarchical density-based assignment.

Area of Science:

• Computational chemistry
• Biophysics
• Molecular dynamics simulations

Background:

• Markov state models (MSMs) are widely used for analyzing molecular dynamics (MD) simulations of biopolymer conformational changes.
• Traditional MSM workflows involve dimensionality reduction, microstate clustering, and macrostate lumping, which can lead to non-Markovian dynamics and require long lag times.

Purpose of the Study:

• To propose a simplified and more effective workflow for constructing MSMs.
• To improve the consistency with the Markovian assumption and accelerate the recovery of slow dynamics.

Main Methods:

• Utilizing hierarchical density-based clustering to naturally separate high-population regions from rarely visited ones in conformational space.
• Implementing a core-set MSM approach based on transitions between metastable states identified by density-based clustering.

Main Results:

• Density-based clustering leads to a more consistent state definition with the Markovian assumption.
• The proposed method effectively recovers the slow dynamics timescales more efficiently compared to traditional methods.
• Validation was performed using a model potential and MD simulations of alanine dipeptide and the FiP35 WW domain.

Conclusions:

• The proposed density-based clustering workflow offers a simplified and more robust approach to building MSMs.
• This method enhances the analysis of conformational transitions in biopolymers by improving Markovianity and timescale convergence.