Faranak Hatami

Asymmetric Focal Loss Improves Graph Neural Network Prediction of Drug-Drug Interactions

Faranak Hatami, Mousa Moradi

Background: Graph neural networks improve computational prediction of polypharmacy side effects, but standard binary cross-entropy training allocates equal capacity to well-classified and difficult examples, potentially missing clinically significant interactions. We evaluated wh…Read more