Proximity Graph Networks (torch_pgn) is a pytorch toolkit allowing for the modular application of multiple different encoder architectures to cheminformatic tasks centered around protein-ligand complexes.
torch-pgn either be installed from PyPi using the pip command or from source. We assume that all users are using conda, if you do not have conda, please install Miniconda from https://conda.io/miniconda.html.
conda create --name torch_pgn python=3.7conda activate torch_pgnpip install torch_pgnconda install pytorch-sparse -c pygconda install -c conda-forge openbabel
Note
If you are using a gpu machine and run into issues with this installation method we suggest you remove pytorch and pyg and reinstall using conda as follows:
conda remove pytorchconda remove pygconda install pytorchconda install pyg -c pygconda install pytorch-sparse -c pyg
git clone https://github.com/keiserlab/torch_pgn/torch_pgn.gitcd torch_pgnconda env create -f environment.ymlconda activate torch_pgnpip install -e