/FGNN

Ligand binding affinity prediction with fusion of graph neural networks and 3D structure-based complex graph

Primary LanguagePython

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.

image

More information is published in the paper.(https://pubs.rsc.org/en/content/articlelanding/2023/cp/d3cp03651k)

Usage of FGNN

After download FGNN, you need to do these firstly:

mkdir data/cache

mkdir data/data_cache

mkdir pdbbind2016/testset

1. Environment

conda env create -f environment-data.yml

conda env create -f environment-model.yml

2. Data preprocessing

conda activate data

python preprocess_pdbbind.py

3. Traing models

conda activate model

python train.py

4. Test and predict

conda activate model

python predict.py

If there are any errors, you may try the code in the 'original version' folder.