DFI-DGCF: A Graph-Based Recommendation Approach for Drug-Food Interactions

Drug discovery focuses on understanding different types of interactions from drug-drug interactions (DDIs) to drug-food interactions (DFIs). The main purpose of DFI is to discover how a particular food affects drug absorption, side effects and its effectiveness. The study of drug-food interactions (DFIs) can provide valuable insight into optimizing patient care, adjusting dosages, and improving patient safety. In this work, we propose a novel workflow where we aim to use a community-based recommender system to infer and identify novel DFIs while incorporating the concept of community profiling and leveraging the power of Graph Neural Networks. We conduct experiments on the DrugBank dataset and use FooDB dataset to learn more about food constituents. Our experiments reveal significant improvements over a number of the latest approaches designed for DFI identification. The findings substantiate that the utilization of multiple graphs to leverage diverse forms of relationships holds the potential to yield better recommendations by extracting complex relationships through the community structure.