Title: State predictive information bottleneck

Abstract

We report the ability to make sense of the massive amounts of high-dimensional data generated from molecular dynamics simulations is heavily dependent on the knowledge of a low-dimensional manifold (parameterized by a reaction coordinate or RC) that typically distinguishes between relevant metastable states, and which captures the relevant slow dynamics of interest. Methods based on machine learning and artificial intelligence have been proposed over the years to deal with learning such low-dimensional manifolds, but they are often criticized for a disconnect from more traditional and physically interpretable approaches. To deal with such concerns, in this work we propose a deep learning based state predictive information bottleneck approach to learn the RC from high-dimensional molecular simulation trajectories. We demonstrate analytically and numerically how the RC learnt in this approach is connected to the committor in chemical physics and can be used to accurately identify transition states. A crucial hyperparameter in this approach is the time delay or how far into the future the algorithm should make predictions about. Through careful comparisons for benchmark systems, we demonstrate that this hyperparameter choice gives useful control over how coarse-grained we want the metastable state classification of the system to be. We thus believe thatmore » this work represents a step forward in systematic application of deep learning based ideas to molecular simulations.« less

Authors:

Wang, Dedi  ^[1];

• Search DOE PAGES for author "Wang, Dedi"
• Search DOE PAGES for ORCID "0000-0002-7524-5809"

Tiwary, Pratyush  ^[1]

• Search DOE PAGES for author "Tiwary, Pratyush"
• Search DOE PAGES for ORCID "0000-0002-2412-6922"

---

+ Show Author Affiliations

1. University of Maryland, College Park, MD (United States)

Publication Date:

Mon Apr 05 00:00:00 EDT 2021

Research Org.:

Univ. of Maryland, College Park, MD (United States)

Sponsoring Org.:

USDOE Office of Science (SC), Basic Energy Sciences (BES); National Science Foundation (NSF)

OSTI Identifier:

1853699

Alternate Identifier(s):

OSTI ID: 1773945; OSTI ID: 1991186

Grant/Contract Number:  

SC0021009; CHE180027P; ACI-1548562

Resource Type:

Accepted Manuscript

Journal Name:

Journal of Chemical Physics

Additional Journal Information:

Journal Volume: 154; Journal Issue: 13; Journal ID: ISSN 0021-9606

Publisher:

American Institute of Physics (AIP)

Country of Publication:

United States

Language:

English

Subject:

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSICS; 37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CHEMISTRY; molecular dynamics; probability theory; transition state; chemical physics; Markov processes; artificial neural networks; artificial intelligence; machine learning; peptides; molecular simulations