FGNN: Ligand binding affinity prediction with fusion of graph neural networks and 3D structure-based complex graph
FGNN is a novel deep fusion graph neural networks framework named FGNN to learn the protein–ligand interactions from the 3D structures of protein–ligand complexes.
More information is published in the paper.(https://pubs.rsc.org/en/content/articlelanding/2023/cp/d3cp03651k)
After download FGNN, you need to do these firstly:
mkdir data/cache
mkdir data/data_cache
mkdir pdbbind2016/testset
conda env create -f environment-data.yml
conda env create -f environment-model.yml
conda activate data
python preprocess_pdbbind.py
conda activate model
python train.py
conda activate model
python predict.py
If there are any errors, you may try the code in the 'original version' folder.