Drug Side Effect Prediction with Deep Learning Molecular Embedding in a Graph-of-Graphs Domain

Drug side effects (DSEs), or adverse drug reactions (ADRs), constitute an important health risk, given the approximately 197,000 annual DSE deaths in Europe alone. Therefore, during development process, detection is of utmost importance, and occurrence ADRs prevents many candidate molecules from going through clinical trials. Thus, early prediction DSEs has potential to massively reduce times costs. In this work, data are represented a non-euclidean manner, form graph-of-graphs domain. such domain, structures molecule by molecular graphs, each which becomes node higher-level graph. latter, nodes stand for drugs genes, arcs represent their relationships. This relational nature represents novelty task, it directly used prediction. For purpose, MolecularGNN model proposed. new classifier based on graph neural networks, connectionist capable processing graphs. The approach improvement over previous method, called DruGNN, as also extracting information graph-based structures, producing task-based fingerprint (NF) adapted specific task. architecture been compared with other GNN models terms performance, showing that proposed very promising.