/BitterGNN

Primary LanguageJupyter NotebookMIT LicenseMIT

BitterGNN

DOI

We created a new GNN (HGNN, combined four modules including edge attention, GAT, MPNN and Weave-GNN), and trained the predictors BitterGNN can distinguish bitter/non-bitter and bitter/sweet taste of molecule.

How to Use HGNN

1. install environment

if use pip (python >= 3.8)
pip install requirements.txt
or use docker
docker build -t my_image .
docker run -it my_image /bin/bash

2. run BitterGNN predictor

python predictor.py -h
python predictor.py -t 0 -m 0 -i test.csv -o result.csv

3. repeat the job of the paper

you can run compare_gnn.ipynb one by one to compare the performance of HGNN with other mainstream GNNs.

you can run compare_taste_predictors.ipynb one by one to compare the performance of BitterGNN predicotrs with other mainstream bitter predicotrs.

Acknowledgments

Supported by Graduate Innovation Fund of Jilin University (Project's number: 2022208)