GradNav: Accelerated Exploration of Potential Energy Surfaces with Gradient-Based Navigation

Introduction

GradNav is an algorithm introduced in our study that enhances the exploration of the potential energy surface (PES), thereby enabling the proper reconstruction of the PES. This approach has been applied to Langevin dynamics within Mueller-type potential energy surfaces and molecular dynamics simulations of the Fs-Peptide protein, demonstrating GradNav's capability to efficiently escape deep energy wells and its reduced reliance on initial conditions. For more detailed information, please refer to our preprint and published paper.

traj_gradnav_muller_200.mp4

Installation

Instructions for the installation will be updated soon. The necessary packages required to run GradNav are listed below:

• openmm version 8.0.0
• scipy version 1.10.0
• mdtraj version 1.9.7

Data

The datasets used in our study are as follows:

• Langevin Dynamics: Simulations of a single particle in Muller-type potentials conducted using OpenMM.
• Molecular Dynamics: Trajectories of Fs-peptide proteins are available at Figshare.

How to Run

All analyses described in the paper are available in the provided Jupyter notebooks:

1. Single Particle LD Simulation in Muller-Type Potential:

   • LD simulation and its analysis: muller_LD.ipynb and modified_muller_LD.ipynb.
   • GradNav application on the Muller potentials case and its analysis: muller_GradNav.ipynb and modified_muller_GradNav.ipynb.
   • Energy surface reconstruction analysis: muller_energy_curve.ipynb.

2. Fs-Peptide MD Simulation:

   • Data for ALA9 and ARG20 used in the paper is located in the data/fspeptide directory.
   • To process your own data, place the original Fs-peptide trajectories in a directory and run:

     python postprocess_peptide-md.py <path_to_md_trajectories> <save_path> --residue <residue_type>
     

   • To apply GradNav in a pseudo-molecular dynamics manner, run run_realsys_GradNav.py.
   • All analysis regarding Fs-peptide can be found in realsys_analysis.ipynb.

Contact

For any inquiries, please reach out to Janghoon Ock at jock@andrew.cmu.edu.

Citation

Please cite our work using the following BibTeX entry:

@misc{ock2024gradnav,
      title={GradNav: Accelerated Exploration of Potential Energy Surfaces with Gradient-Based Navigation}, 
      author={Janghoon Ock and Parisa Mollaei and Amir Barati Farimani},
      year={2024},
      eprint={2403.10358},
      archivePrefix={arXiv},
      primaryClass={physics.chem-ph}
}