D4CMPP 0.21

Molecular property prediction based on Graph Convolution Network published by Deep4Chem

D4C molecular property prediction

This project is a deep learning application designed to predict molecular properties. The models implemented in this project feature interpretable and hierarchical architectures, including conventional graph convolutional models.

How to start

1. Place the CSV file for training in the "_Data" folder.
   • The SMILES strings of molecules should be in the "compound" column.
   • There needs to be at least one molecular property for each corresponding molecule.
2. Check and choose the ID of the deep learning model in "network_refer.yaml".
3. Run the "main.py" file as

python main.py -n [network_id] -d [data_file_name] -t [property_column]

For example,

python main.py -n GCN -d Aqsoldb -t Solubility

The graph cache file will be saved in "_Graph" folder. The trained result will be saved in "_Model" folder.

Code Attribution and Licensing Information

This project includes code from the GC-GNN by Adem Rosenkvist Nielsen Aouichaoui ([e-mail]), licensed under the MIT License. https://github.com/gsi-lab/GC-GNN/tree/main