Discussing Markov state models after two decades of development

This repository is the home of a LiveCoMS Roundtable article on Markov state models (MSMs). Like other living documents in the Living Journal of Computational Molecular Science (LiveCoMS), it is maintained openly on GitHub so that it can be updated, corrected, and improved over time.

Markov state models play a central role in the analysis of molecular dynamics (MD) simulations of biomolecules, and their methodology has been under continual development since their introduction in the early 2000s. MSMs take MD trajectory data as input and produce a compact, coarse-grained, and interpretable representation of the system in terms of conformational states and quantitative observables such as equilibrium populations and interconversion timescales. Despite two decades of progress there is still no consensus on best practices, on the limitations of MSMs, or on whether the first-order Markovian framework should be extended or abandoned.

This Roundtable gathers diverse opinions from more than ten authors active in MSM development. By design, each author or author group offers a separate contribution and no consensus view is imposed. The article is meant as a guide for readers who wish to be exposed to a wide set of expert perspectives.

The Paper

Versions

The paper is written and maintained here on GitHub. Tagged releases correspond to the versions submitted for peer review and, once published, archived with a DOI. The version number is set in MSM-Roundtable/main.tex via the \versionnumber macro.

Publication in LiveCoMS

This document uses the LiveCoMS roundtable article class. Roundtable articles are a pilot format in which several authors each contribute an independent perspective on a shared topic. As with all LiveCoMS articles, it is subject to peer review.

Ongoing updates and credit

MSM methodology continues to evolve, and this document is intended to evolve with it. Community input is welcome. Suggestions, corrections, and requests for clarification can be raised through the GitHub issue tracker, and contributions that improve the article will be acknowledged.

Citing this work

Please cite the published LiveCoMS version once a DOI has been assigned. Citation details (DOI, volume, issue, year) will be added here on publication. Until then, please cite this repository and the relevant version tag.

The Vision

MSM building is often described as something of an art. A wide range of software, mathematical formalism, and machine-learning tools now exists, which can make it overwhelming to choose the best strategy for extracting equilibrium and kinetic information from simulations. Excellent textbooks and review articles exist, but experts still disagree on important points.

The aim of this Roundtable is to expose beginners and practitioners alike to the current state of the art, including its criticisms and alternatives, by placing a diverse set of expert viewpoints side by side. Rather than prescribing a single recommended workflow, the article lets readers compare perspectives and form their own judgement across the full pipeline, from data generation to model interpretation.

Contents

The article opens with an Introduction and a shared Glossary of terms and symbols, followed by independent perspectives:

1. Fine-grained models for quantitative prediction of experiments — Bowman
2. Influence of limited sampling on MSM and other methods — Caflisch and Giangreco
3. Computing kinetic statistics with MSMs — Dinner and Weare
4. From projection operators to Markov state models: features, transition states, and memory — O'Connor and Huang
5. From operator learning to molecular modeling — Keller
6. Koopman modeling of molecular kinetics — Klus, Mey, and Nüske
7. Objective choice controls optimal biasing on MSM networks — Rosta and Fadhluddin
8. MSMs, collective variables, and effective dynamics — Schütte
9. Feature selection via MoSAIC contact clusters — Stock
10. VAMP is not a panacea — Voelz
11. Two matrices to rule them all — Zuckerman

The article closes with Concluding Remarks.

Get Involved

Feedback and contributions are welcome:

• Raise an issue to point out errors, suggest clarifications, or propose topics that a future version should cover.
• Open a pull request to fix typos, improve figures, or extend a perspective (please coordinate with the relevant author for substantive changes).

Authors and Acknowledgments

Coordinating authors: Daniel M. Zuckerman (Oregon Health and Science University) and Antonia S. J. S. Mey (University of Edinburgh).

Contributing authors (alphabetically): Gregory R. Bowman, Amedeo Caflisch, Aaron R. Dinner, Ahmed Fadhluddin, Giuseppe Giangreco, Xuhui Huang, Bettina G. Keller, Stefan Klus, Antonia S. J. S. Mey, Feliks Nüske, Michael O'Connor, Edina Rosta, Christof Schütte, Gerhard Stock, Vincent Voelz, Jonathan Weare, and Daniel M. Zuckerman.

We thank the LiveCoMS editors for supporting this pilot Roundtable format, and the wider MSM community whose feedback helps keep this document current.

Manifest

.
├── README.md                     This file
└── MSM-Roundtable/
    ├── main.tex                  Main LaTeX source for the article
    ├── refs_msm_roundtable.bib   Bibliography
    ├── livecoms.cls              LiveCoMS document class
    ├── vancouver-livecoms.bst    Bibliography style
    ├── StockFigureMSM.pdf        Figure used in the article
    └── figures/                  Figures for the individual perspectives

Licensing

The manuscript text and figures are made available for community use. A specific licence for this repository will be stated here; LiveCoMS articles are typically released under a Creative Commons licence (CC-BY), with associated code, where present, under a permissive open-source licence.