Generating Protein Folding Trajectories Using Contact-Map-Driven Directed Walks.

Area of Science:

• Computational Biology
• Biophysics
• Machine Learning

Background:

• Accurate prediction of protein structure has advanced with machine learning.
• However, generating and characterizing protein-folding pathways remains a significant challenge.
• Traditional methods often require long simulation times and specific order parameters.

Purpose of the Study:

• To develop a new computational strategy for generating protein-folding trajectories.
• To enable thermodynamic and kinetic characterization of protein-folding paths.
• To overcome limitations of existing protein-folding simulation techniques.

Main Methods:

• Employed a directed walk strategy in residue-level contact-map space.
• Viewed protein folding as discrete transitions between potential energy surface minima.
• Utilized reaction-path analysis for thermodynamic and kinetic characterization.

Main Results:

• Successfully generated protein folding trajectories using the discretized-walk strategy.
• Validated paths against molecular dynamics simulations for coarse-grained proteins.
• Demonstrated that ranking paths by energy barriers identifies physically sensible folding ensembles.

Conclusions:

• The directed walk approach in contact-map space circumvents challenges of long time scales and order parameter selection.
• This method offers a computationally efficient route for studying protein folding.
• Provides a valuable new tool for understanding the complex protein-folding problem.