This repository is the version with minor changes to molecularGNN_smiles.The original algorithm for obtaining atom IDs based on r-radius subgraphs has been retained, the preprocessing convenience has been improved. Specifically, changes have been made in the following three areas.

• The SMILES input format has been modified to support csv.
• Creating Featurizer class, we have made it easier to manipulate the molecular graph fingerprint acquisition for GNN from SMILES.
• The model validation process has been modified. In the preprocessing phase, the training dataset is partitioned based on k-fold cross-validation to generate subsets of k different training and validation sets. During training of the GNN, the test set predictions of each model built from each subset are saved.