A playground for applying graph convolutional networks to molecules, with a focus on learning continuous "atom-type" embeddings and from these classical molecule force field parameters.
This framework is mostly based upon the End-to-End Differentiable Molecular Mechanics Force Field Construction preprint by Wang, Fass and Chodera.
Examples for using this framework can be found in the examples directory.
The required dependencies for this framework can be installed using conda:
conda install -c conda-forge -c dglteam nagl "dgl >=0.7"
To make the full use of the framework, it is recommended to install the following
# Molecule labelling
conda install -c conda-forge openff-toolkit
The main package is release under the MIT license. Parts of the package are inspired by / modified from a number of third party packages whose licenses are included in the 3rd party license file.
Copyright (c) 2021, Simon Boothroyd