Coloring Molecules with Explainable Artificial Intelligence for Preclinical Relevance Assessment

Graph neural networks are able to solve certain drug discovery tasks such as molecular property prediction and de novo molecule generation. However, these models considered “black-box” “hard-to-debug”. This study aimed improve modeling transparency for rational design by applying the integrated gradients explainable artificial intelligence (XAI) approach graph network models. Models were trained predicting plasma protein binding, hERG channel inhibition, passive permeability, cytochrome P450 inhibition. The proposed methodology highlighted features structural elements that in agreement with known pharmacophore motifs, correctly identified cliffs, provided insights into unspecific ligand–target interactions. developed XAI is fully open-sourced can be used practitioners train new on other clinically relevant endpoints.