This repository contains code and examples for the article: “Rapid Generation of Rare Event Pathways Using Direction-Guided Adaptive Sampling: From Ligand Unbinding to Protein (Un)Folding.” 📄 DOI: https://doi.org/10.1021/acs.jctc.5c01244

PathGennie is a general-purpose steering framework for guiding molecular simulations along data-driven(or physical) collective variables (CVs) to rapidly sample rare event transitions such as:

• Ligand unbinding
• Protein folding and unfolding

It leverages high-performance libraries like OpenMM and MDAnalysis, and includes tooling for CV construction, adaptive sampling, and optimized trajectory generation.

pathgennie/ │ ├── README.md # Documentation and usage guide ├── LICENSE # MIT or compatible license ├── environment.yml # Conda environment for reproducibility │ ├── Scripts/ # Scripts for various path generation tasks │ ├── unbind # Ligand unbinding module │ ├── unfold # Protein unfolding module │ └── fold # Protein folding / reverse folding module │ ├── examples/ # example systems │ ├── 3PTB/ # Example: Bovine Trypsin Inhibitor │ │ ├── native.pdb │ │ ├── start.gro │ │ ├── topol.top │ │ └── system.py │ │ │ └── 2JOF/ # Example: Trp-cage protein system │ ├── native.pdb │ ├── start.gro │ ├── topol.top │ └── system.py 

Clone the repository and set up the Conda environment:

# Clone the repository git clone https://github.com/dmighty007/PathGennie.git # Navigate to the project folder cd PathGennie # Create and activate the environment conda env create -f environment.yml conda activate pathgennie

This installs all required dependencies, including:

• openmm
• mdanalysis
• numpy, numba, tqdm, and more

✅ Note: Ensure Miniconda or Anaconda is installed before proceeding.

This framework enables ligand unbinding simulations by steering MD along principal components derived from distance features between a ligand and its binding site. These components form a low-dimensional CV space for guided sampling.

1. Create a working directory:

   mkdir Test && cd Test

2. Add your input files:

   pbcmol.gro # Structure with solvent and PBC topol.top # GROMACS-compatible topology 

3. Perform energy minimization and equilibration using GROMACS or OpenMM. These outputs will be used as initial configurations.

Generate a PCA model based on ligand–protein contacts:

python pcagen.py pbcmol.gro \ --ligand_sel "resname LIG" \ --output pca.pkl

This script:

• Computes distance features between the ligand and surrounding protein atoms
• Performs PCA on the feature matrix
• Stores the principal components in pca.pkl

Create a system.py file to build the OpenMM simulation system:

from openmm.app import * from openmm import * from openmm.unit import * class Simulation_obj: def __init__(self): gro = GromacsGroFile('pbcmol.gro') top = GromacsTopFile('topol.top', periodicBoxVectors=gro.getPeriodicBoxVectors(), includeDir='/usr/local/gromacs/share/gromacs/top') system = top.createSystem(nonbondedMethod=PME, nonbondedCutoff=1*nanometer, constraints=HBonds) integrator = LangevinMiddleIntegrator(300*kelvin, 1/picosecond, 0.004*picoseconds) self.simulation = Simulation(top.topology, system, integrator)

📘 Refer to the OpenMM documentation for advanced setup options.

Run the unbinding driver with:

../../Scripts/unbind \ --structure_file pbcmol.gro \ --verbose \ --relax1 10 \ --relax2 15 \ --max_probes 50 \ --temperature 300 \ --model_file pca.pkl

Output:

• trajectory.xtc: Reactive trajectory file showing unbinding progression

💡 For full options, run:

../../Scripts/unbind -h

The same logic applies to protein folding/unfolding simulations.

../../Scripts/unfold \ --ref_config folded.gro \ --start_config folded.gro \ --verbose \ --relax1 10 \ --relax2 15 \ --max_probes 50 \ --temperature 290

../../Scripts/fold \ --ref_config folded.gro \ --start_config equili.gro \ --verbose \ --relax1 10 \ --relax2 15 \ --max_probes 50 \ --temperature 290

📌 Make sure system.py is present in the working directory for these commands.

The examples/ directory includes two protein systems with:

• Native PDB structure
• Starting gro configuration
• GROMACS topology (topol.top)
• Compatible system.py

You can directly run folding/unfolding or unbinding simulations on these examples.

This project is licensed under the MIT License.

Feel free to open an issue or submit a pull request!