A Graph Convolutional Deep Learning Model for Identifying Potential Aromatase Inhibitors

Aromatase inhibitors play a crucial role in the treatment of hormone-sensitive breast and lung cancers by blocking the conversion of androgens to estrogens. However, there is a need for more effective and selective aromatase inhibitors. Recent advancements in deep learning devoted to drug discovery are the application of graph convolutional neural networks to predict the properties of small molecules represented as graphs. In this study, we utilize the inhibitory activities of experimentally tested compounds against aromatase from the PubChem database. We develop a graph convolutional deep neural network capable of predicting molecules with aromatase inhibitory properties. The model with the most optimal hyperparameter set exhibits an AUC of 0.83. Our model serves as a good starting point for further development and for screening potential aromatase inhibitors, which can help save time and resources that would otherwise be spent on trial-and-error experiments.