AtomiGraph

AtomiGraph (from Atomistic Graph) is a Python tool for transforming bonded topologies from molecular dynamics (MD) simulations into graph representations using NetworkX.

It can be used to analyze bonded network structures in detail or to identify changes in the covalent bond network (i.e., reactions) over time.

This started as a side project while finishing my PhD at IMWF, University of Stuttgart — so parts of the code are still a bit rough and not fully optimized or feature-complete (e.g., automated kinetics or reaction rates).

That said, it already provides a useful framework for exploring topology and reaction mechanisms in MD simulations.

Disclaimer:
Provided as is, without any warranty. Use at your own risk — but feel free to use, modify, and build on it. Contributions are very welcome.

---

What AtomiGraph does

• Parses LAMMPS data files and ReaxFF bond topology dumps
• Transforms bonded topology into NetworkX graph representations
• Enables direct application of NetworkX algorithms and graph-theoretical analyses
• Supports removal of nodes by atom type or pattern (e.g., for cleanup or coarse-graining)
• Identifies reaction events by tracking connectivity changes between timesteps
• Filters out reversible reactions within a defined time window
  (e.g. A + B → C followed by C → A + B)
• Generates before/after visualizations for detected reactions

---

What AtomiGraph does not (yet)

• Does not consider atomic positions or geometry
• Does not include non-covalent interactions (e.g. hydrogen bonds, ionic interactions)
• Does not generate SMILES / SMARTS or other cheminformatics outputs
• Does not write out coarse-grained MD configurations (yet)
• No automated extraction of reaction rates (yet)

---

Citing

No formal publication yet.

If you use AtomiGraph in academic work, please cite:

Wolfgang Verestek, AtomiGraph (GitHub repository)

A methods paper is currently in preparation.

---

Prerequisites

Python 3 and the following modules:

• sys, os, random
• numpy
• pandas
• matplotlib
• networkx

---

Installation

Add the AtomiGraph base folder to your PYTHONPATH, e.g.:

export PYTHONPATH=$PYTHONPATH:/path/to/AtomiGraph

---

Usage

import atomigraph as ag

net = ag.AtomiGraph(
    infile="bonds.reaxff.dump",
    atom_type_map="1:C,2:H,3:H,4:O,5:O,6:O,7:O,8:O"
)

net.read()
net.find_rxns()
net.plot_rxns()

---

Command line usage:

tbd

---

Frame/Reaction sampling

Frame comparison can be controlled via:

• startstep: starting frame index
• checkstep: compare frame i up to i + checkstep
• framestep: increment between evaluations

Example: startstep = 0 checkstep = 1 framestep = 5

-> compares: 0 vs 1, 5 vs 6, 10 vs 11, ...

---

Visualization

Color coding can be specified for each atom type in the dictionary "type2color" in utils.py.

Default colors follow Jmol conventions: https://jmol.sourceforge.net/jscolors/

Contributing

Contributions, ideas, and improvements are very welcome.

This started as a small playground project — feel free to help shape where it goes next.