xchimes

Repository for the paper "A Generalized Machine-Learning Framework for Developing Alchemical Many-Body Interaction Models for Polymer Grafted Nanoparticles"

We present the source code to perform Forward-Reverse (FR) method for potential of mean force (PMF) generation using HOOMD-blue and to use ChIMES-Calulater in HOOMD-blue to conduct coarse-grained (CG) molecular dynamics simulation.

Related database generated using the code present in this repository: http://deepblue.lib.umich.edu/data/concern/data_sets/v979v4270.

Prerequisites

The source code requires the following packages:

• HOOMD-blue
• ChIMES-Calculator
• freud
• gsd
• Numpy
• matplotlib

Usage

We explain the usage of three folders:

1. fr_method:

   The Jupyter notebooks, fr_demo_2b.ipynb and fr_demo_3b.ipynb demonstrate the example and procedure to calculate two- and three-body PMF using the principal of FR method, where we use the Lennard-Jones particle as example for simpilicity. The steered molecular dynamics (SMD) method is used to calculate the work along reaction coordinate.

   After running the two notebooks, it will generate two output for each notebook, including the SMD results: smd.txt and smd_3b.txt, as well as the SMD trajectories: traj.gds and traj_3b.gsd.

   To calculate the three-body PMF, an interpolation method must be used to obtain the surrogate model for the two-body PMF and conduct the three-body PMF calculation. Here, we use the ChIMES model as interpolation method.

2. hoomd-chimes:

   The OS-CG_PGN_sim.py demonstrates an example of using ChIMES-Calculator to simulation the PGNs with $\alpha=4$ in hoomd. Notice that user need to input the right path pointing to ChIMES-Calculator in the script.

   The hoomd_chimes_addons.py glue the ChIMES-Calculator with HOOMD-blue together by using the custom force module.

   The expected output files are also included.

3. src:

   The smd.py implements the two- and three-body SMD modules using hoomd's custom force module.

   The utilities.py implements a subset of ChIMES-LSQ to aid the PMF calculation.

Cite this work

1. Zhang MY, Lee S-K (Alex), Glotzer S, Lindsey R. A Generalized Machine-Learning Framework for Developing Alchemical Many-Body Interaction Models for Polymer Grafted Nanoparticles. ChemRxiv. 2025; doi:10.26434/chemrxiv-2025-n4d3s