A substantial research effort has gone into the development and improvement of kinetic modeling tools for the analysis of time-series data, particularly molecular dynamics simulations. For example, Markov state models (MSMs) offer a master equation framework that can be variationally optimized—and thus automated—using machine learning techniques. However, the theoretical underpinnings of the MSM approach require that the dynamics under investigation be reversible—i.e., in thermodynamic equilibrium. Because many interesting phenomena in chemistry and related fields cannot be accurately approximated as reversible, MSMs are fundamentally limited in their applicability. Here, we showcase and discuss the advantages and disadvantages of two algorithmic avenues for the kinetic modeling of irreversible systems, both recently adapted from computational fluid dynamics. First, the Koopman model framework, which is already gaining traction in the computational chemistry and molecular biology communities, represents a generalization of MSMs to nonequilibrium systems, albeit with some loss of interpretability. Second, the identification of coherent structures in turbulent fluid flows has been very recently shown be applicable to molecular systems. In addition to describing the state of art in irreversible kinetic modeling, we present new algorithms for both Koopman modeling and coherent structure identification in the context of molecular dynamics data analysis with emphases on interpretability and connections to experimental observables.

Extending molecular kinetics modeling tools to nonreversible dynamics

2019 Fall National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.