Official implementation of MGIB model in the paper: Self-Supervised Pre-Training via Multi-View Graph Information Bottleneck for Molecular Property Prediction (IEEE Journal of Biomedical and Health Informatics 2024) [MGIB]

conda env create -f environment.yaml

Download the pre-training dataset from [ZINC15(2M)] and place "zinc_combined_apr_8_2019.csv" in the "./data/pretrain/zinc_2m.csv".

You can pretrain the model by

python main_pretrain.py

Download the downstream dataset from [MoleculeNet]. Place the raw data in CSV format in the "data/down_task/dataset_name/raw/dataset_name.csv, for example "./data/down_task/sider/raw/sider.csv".

You can finetune the model for classification tasks by

python main_clr_downstream.py

Please kindly cite this paper as follows. Thank you.

@article{zang2024self,
  title={Self-supervised pre-training via multi-view graph information bottleneck for molecular property prediction},
  author={Zang, Xuan and Zhang, Junjie and Tang, Buzhou},
  journal={IEEE Journal of Biomedical and Health Informatics},
  volume={28},
  number={12},
  pages={7659--7669},
  year={2024},
  publisher={IEEE}
}