Predicting-CO2-Absorption-in-Ionic-Liquid-with-Molecular-Descriptors-and-Explainable-GNN

Data and code for Predicting $CO_2$ Absorption in Ionic Liquid with Molecular Descriptors and Explainable GNN

How to run the code

reproducing the result for predicting properties with shallow machine learning

• Entering /Shallow_Machine_Learning_for_property_prediction, each jupyter notebook contain a reproducing code for each type of shallow machine learning method mentioned in paper

reproducing the result for predicting properties with GNN

• Entering /GNN_for_property_prediction
• run python GIN_Runner.py to reproduce GIN model result
• run python GAT_Runner.py to reproduce GAT model result
• run python GCN_Runner.py to reproduce GCN model result
• noted that accuracy on test dataset may vary a little bit each time you run the code due to the random spliting for train and test dataset

reproducing the result for fragment importance explanation with GNN Explainer

• Entering /Explainer_for_ionic_molecule
• run explain_whole_dataset.ipynb to visualizing the fragment importance explanation for the whole dataset
• run explain_single_ionic_molecule_pair.ipynb to visualizing the explanation for single ionic molecule pair in hotmap form
• run python fragment_explain.py to reproduce the fragment importance explanation process for the whole dataset

About the data

• Due to the reason that the original data has been adapted into different form for different task, we separately clean up a dataset with both smiles dictionary and whole dataset and store those file in Original_Dataset

Reference

If you find the code useful for your research, please consider citing

@inproceedings{
  Yue2022predictCO2,
  title={Predicting CO2 Absorption in Ionic Liquid with Molecular Descriptors and Explainable Graph Neural Networks},
  author={Yue Jian and Yuyang Wang and Amir Barati Farimani},
  booktitle={},
  year={2022},
  url={}
}

This work is built upon some previous papers and opensource package which might also interest you:

• Fey, Matthias and Lenssen, Jan E. "Fast Graph Representation Learning with PyTorch Geometric" ICLR Workshop on Representation Learning on Graphs and Manifolds. 2019.
• Zhitao Ying, Dylan Bourgeois, Jiaxuan You, Marinka Zitnik, Jure Leskovec. "GNNExplainer: Generating Explanations for Graph Neural Networks" Advances in Neural Information Processing Systems. 2019.