Determination of reaction coordinates via locally scaled diffusion map.

Area of Science:

• Computational chemistry and biophysics
• Data analysis and machine learning applied to molecular dynamics

Background:

• Understanding macromolecular dynamics is crucial for molecular biology.
• Existing methods struggle to capture collective motions and their associated time scales effectively.

Purpose of the Study:

• To develop a novel multiscale method for determining collective reaction coordinates in macromolecular dynamics.
• To correlate these coordinates with the inherent time scales of molecular motion.
• To provide a generalizable technique applicable to various molecular systems.

Main Methods:

• Utilizing diffusion maps and local intrinsic dimensionality analysis for large datasets.
• Accounting for local variations within molecular configuration space.
• Applying the method to all-atom alanine dipeptide and coarse-grained src homology 3 protein domain.

Main Results:

• Generated global coordinates that correlate with molecular motion time scales.
• Provided clear physical interpretations for the derived collective coordinates.
• Successfully calculated transition rates using the new coordinate system.

Conclusions:

• The developed multiscale method offers a robust approach for analyzing macromolecular dynamics.
• The technique provides physically interpretable reaction coordinates linked to motion timescales.
• This method is broadly applicable to any system with available Boltzmann-sampled configurations.