Markov state models revisited: Principles and algorithms for unbiased observables
David Aristof, Robert J. Webber, Daniel M. Zuckerman
arXiv:2607.19452·cond-mat.stat-mech·Published 2026-07-21
Markov state models (MSMs) have become ubiquitous tools for analyzing molecular dynamics (MD) simulations because of their simple, powerful premise: although complete MD sampling may be impossible, the MSM can "stitch together" transition probabilities derived from local sampling to provide a global picture of kinetics and mechanisms. In the standard MSM framework, the available MD data is organized into a single transition matrix, which is then used to estimate all observables at a lag time chosen so the coarse-grained dynamics are approximately Markovian. This approach leads to avoidable model bias and motivates long lag times that obscure short-timescale processes of interest. In contrast, this paper shows how to obtain unbiased coarse-grained observables at any fixed lag time and for any fixed coarse-graining in the limit of infinite, properly weighted data. The central idea is to replace the single-matrix framework with two transition matrices -- one representing equilibrium dynamics and another representing source-sink recycling dynamics -- and use the correct matrix or matrices to estimate the matched dynamical observables.
TopicsMolecular Dynamics & Multiscale Modeling
Tagscoarse-grained molecular-dynamics
arXiv categoriescond-mat.stat-mech, physics.chem-ph, q-bio.BM
arXiv abstract pagePDF