A Data-Driven Perspective on the Hierarchical Assembly of Molecular Structures.

Area of Science:

• Computational Biology
• Biophysics
• Statistical Mechanics

Background:

• Macromolecular systems possess numerous atomic degrees of freedom, posing challenges for simulating long-timescale dynamics.
• Existing coarse-graining methods for biomolecules lack a systematic approach for defining optimal representations.
• Understanding large-scale structural changes requires models that simplify atomic details.

Purpose of the Study:

• To develop a systematic approach for learning coarse-grained representations of macromolecules from microscopic simulation data.
• To identify effective coarse variables by partitioning degrees of freedom in both physical and conformational spaces.
• To enable a multiscale description of macromolecular systems in space and time.

Main Methods:

• Partitioning atomic degrees of freedom in structural and conformational spaces.
• Identifying dynamically coherent groups of particles within metastable states.
• Applying the learned coarse-grained models to protein folding dynamics simulations.

Main Results:

• A systematic method for defining coarse variables and learning coarse-grained models was developed.
• The approach revealed a multiscale description of macromolecular dynamics.
• Analysis of protein folding dynamics provided a revised perspective on prestructured regions (foldons).

Conclusions:

• The proposed systematic coarse-graining approach effectively captures essential dynamics of macromolecular systems.
• This method facilitates a multiscale understanding of protein folding and hierarchical structure assembly.
• The findings offer a new framework for studying complex biological processes at reduced resolution.