maintainers:	andrewtarzia
documentation:	https://cgexplore.readthedocs.io/en/latest/

Overview

cgexplore or cgx is a general toolkit built on stk for constructing, optimising and exploring molecular coarse-grained models.

Important

Warning: This package is still very much underdevelopment and many changes are expected.

In particular, if you are trying to reproduce exactly our recent work on structure prediction, then you should install an old version of the code (cgexplore==2025.2.5.1) alongside the topology_scrambler code and use these docs. Note, however, that the recipes actually reproduce that work with the updated interface.

Installation

cgexplore can be installed with pip:

pip install cgexplore

Then, update directory structure in env_set.py if using example code.

The library implements some analysis that uses Shape 2.1. Follow the instructions to download and installed at Shape

Developer Setup

To develop with cgexplore, you can clone the repo and use just and uv to setup the dev environment:

just setup

Usage

We are moving toward implementing a recipe list, which can be found in the recipe page.

To reproduce data in DOI: `10.1039/D3SC03991A <https://doi.org/10.1039/D3SC03991A>`_: Download the source code from first_paper_example - presubmission release from Releases.I do not guarantee that running the example code on the current version will work. However, with each pull request a test is run as a GitHub Action connected to this repository. This ensures that the results obtained for a subset of the original data set do not change with changes to this library. Additionally, the naming convention has changed and force field xml files should provide the appropriate information for mapping angles to models.

• The directory cgexplore contains the actual source code for the package.
• The directory first_paper_example contains the code for 10.1039/D3SC03991A.
  • generate_XX.py generates cage structures for different topology sets
  • env_set.py sets a specific environment for file outputs
  • plot_XX.py produces images and figures, and performs analysis

Important:

Warning: If you have a CUDA-capable GPU and attempt to use CUDA in the first example, you may get NaN errors due to the torsion restriction for angles at 180 degrees, which cause problematic forces. This will be handled in future versions of the code. And logically, I would suggest removing the torsion restriction for those angles. The platform can be handled through this argument in build_building_blocks and build_populations, which I currently set to None, meaning OpenMM will decide for itself.

How To Cite

If you use cgexplore, please cite this paper

https://doi.org/10.26434/chemrxiv-2025-f034c

and reference this URL

https://github.com/andrewtarzia/CGExplore

If you use our minimial model, please cite this paper

Systematic exploration of accessible topologies of cage molecules via minimalistic models

Publications using CGExplore

• Using stk for constructing larger numbers of coarse-grained models: Systematic exploration of accessible topologies of cage molecules via minimalistic models.
• Starship structure prediction: (Adjacent backbone interactions control self-sorting of chiral heteroleptic Pd3A2B4 isosceles triangles and Pd4A4C4 pseudo-tetrahedra)
• Structure prediction: (Predicting stable cage structures by enumerating stoichiometry and topology)

Acknowledgements

Funded by the European Union - Next Generation EU, Mission 4 Component 1 CUP E13C22002930006.

This work is now developed as part of the Tarzia Research Group at the University of Birmingham.