puraaaa / protein-ligand-interaction-graphs

Protein-Ligand-Interaction-Graphs

This is the repository for https://www.biorxiv.org/content/10.1101/2022.03.04.483012v1?rss=1

Python Packages

Install the needed python packages using conda with the following command:

conda env create -f torch_geo.yml

Alternatively, the following packages can be install manually (not preferred). Please make sure the correct version of PyTorch is installed!!! Other versions of PyTorch (eg. 1.10) will crash the models.

• conda create --name torch_geo
• conda install -c conda-forge rdkit=2021
• conda install pyyaml
• conda install pytorch=1.9.0 cpuonly -c pytorch
• conda install pytorch-geometric -c rusty1s -c conda-forge
• conda install -c conda-forge optuna
• conda install -c conda-forge gpytorch
• conda install -c conda-forge tqdm

1) Create Protein-Ligand Interaction Graphs

The code needed to generate PLIGs from a protein-ligand complex can be found in the "PLIG_tutorial/" folder.

2) Run a PLIG, ligand-based GNN or MLPNet model

All GNN implementations of PLIGs, ligand-based GNNs, as well as the MLPNet implementation of ECFP/FCFP and ECIF fingperints can be found in the "models_main/" folder.

The following data is supplied:

i) All hyperparameter tuned GNN PLIG, GNN ligand-based, MLPNet ECIF and MLPNet ECFP/FCFP models

ii) Pre-prepared features (PLIGs, ECIF, ECFP/FCFP) for the PDBbind 2020 general + PDBbind 2016 refined set

iii) All config files needed to train all model+feature combinations on crystal and docked poses.

3) Figures and data presented in the publication

All raw-data to generate the figure published in https://www.biorxiv.org/content/10.1101/2022.03.04.483012v1?rss=1 can be found in the "publication" folder