Commute Maps: Separating Slowly Mixing Molecular Configurations for Kinetic Modeling

Area of Science:

• Computational Chemistry
• Molecular Dynamics
• Biophysics

Background:

• Accurate kinetic models of macromolecular systems require distinguishing slowly interconverting states.
• Measuring distances between molecular configurations is essential for this task.

Purpose of the Study:

• To define a novel distance metric, the commute distance, for molecular configurations.
• To demonstrate its utility in building kinetic models from molecular dynamics data.

Main Methods:

• Definition of commute distance, related to expected commute time between configurations.
• Approximation of commute distance using Markov operator eigenfunctions from variational approach of conformation dynamics (VAC) or time-lagged independent component analysis (TICA).
• Scaling VAC/TICA components to create a commute map where Euclidean distance equals commute distance.

Main Results:

• Commute distance provides a robust measure for distinguishing molecular states.
• The commute map enables kinetic model computation using standard Euclidean operations like clustering.
• Introduction of 'total kinetic content' as a metric for ranking feature sets and model quality.

Conclusions:

• Commute distance offers a practical and effective method for analyzing molecular dynamics data.
• This approach facilitates the construction of accurate kinetic models, such as Markov state models.
• The total kinetic content provides a valuable tool for optimizing feature selection and model assessment.