Random walks in a free energy landscape combining augmented molecular dynamics simulations with a dynamic graph...

Area of Science:

• Computational Chemistry
• Biophysics
• Molecular Dynamics

Background:

• Understanding protein folding/unfolding mechanisms is crucial for molecular biology.
• Conformational search in molecular dynamics (MD) simulations is computationally intensive.
• Characterizing transition states requires advanced simulation techniques.

Purpose of the Study:

• To enhance conformational search for determining transition states in potential energy surfaces.
• To investigate peptide folding/unfolding phase transitions using novel MD approaches.
• To elucidate the role of water in protein denaturation processes.

Main Methods:

• Augmented dynamics with swarm particle intelligence and Tsallis statistics MD simulations.
• Dynamic graph neural networks to model free energy landscape dynamics.
• Multi-digraph approach for identifying discrete conformational transition pathways.
• Information-theoretic measures like local transfer entropy to analyze peptide-solvent interactions.

Main Results:

• The study successfully mapped transition states and pathways for peptide folding/unfolding.
• Dynamic graph neural networks effectively modeled the evolution of free energy states.
• Water's role in denaturation was clarified: folding reduces 'iceberg' formation, while unfolding shifts equilibrium towards it.
• Dipole-dipole correlations quantified via transfer entropy explain water's influence on folding/unfolding.

Conclusions:

• Novel augmented dynamics and graph-based methods enhance conformational search accuracy.
• The findings provide a detailed molecular-level understanding of water's role in protein denaturation.
• This work offers insights into designing peptides with specific folding behaviors.