Data-driven computation of molecular reaction coordinates.

Area of Science:

• Computational Chemistry
• Molecular Dynamics
• Theoretical Chemistry

Background:

• Identifying reaction coordinates is crucial for understanding complex molecular systems with slow dynamics.
• Recent work defined reaction coordinates via a transition manifold theory.
• This theory enabled a numerical method using short parallel molecular dynamics (MD) simulations.

Purpose of the Study:

• To extend the transition manifold theory for practical computational chemistry applications.
• To develop a computational scheme for global reaction coordinate computation.
• To link new reaction coordinates to transition path theory and Markov state models.

Main Methods:

• Developed an alternative computational scheme for global reaction coordinate calculation.
• Utilized single long molecular trajectories or push-forward of point clouds as input data.
• Employed a Galerkin approximation of transition manifold reaction coordinates, tunable via ansatz functions.

Main Results:

• Proposed a ready-to-implement variant using data-fitted, mesh-free ansatz functions.
• Demonstrated the method's efficacy on a small protein system.
• Enabled global computation of reaction coordinates from standard simulation data.

Conclusions:

• The extended method enhances the practical applicability of transition manifold theory in computational chemistry.
• The approach provides accurate approximations of long-time system behavior.
• This work offers a versatile tool for analyzing complex molecular dynamics.