Graph neural network to assign bond vibrational frequency

Neural network model discussed in "A graph-based machine learning framework to assign empirical interaction parameters for novel molecules" (doi)

It is recommeded that you use a new conda environment to install this package and its dependencies.

conda create --name gravy python=3.11.8

Dependencies

• python 3.11.8
• cuda 12.1 (optional, see point about DGL)
• chemical_equivalence
• atb_output
• NXMol
• DGL If you are not on Linux and/or not on a NVIDIA GPU, you need to install DGL manually. Ensure that you choose a version compatible with torch 2.4.x as that is the version gravy is tested on. Otherwise, proceed with installation instructions below.

Installation

# clone this repo
cd Gravy
pip install .

Inference

Example PDB files are in src/gravy/examples. To execute the dexverapamil example, simply run python query.py.

To use your own PDB file (geometry should be optimised), edit query.py so that

PDB_PATH
MOL_NAME
NET_CHARGE

reflect your molecule of interest.

TODO

• Manual passing of fractional bond orders during PDB