relevant. Artificial intelligence is increasingly integrated into scientific discovery, but two key challenges remain: the integration of domain knowledge into models and the generation of explainable outputs that are scientifically This talk explores the role of knowledge graphs in addressing these challenges through the lens of biomedical research with a particular focus on protein-protein interaction prediction and drug repurposing. It highlights the limitations of current knowledge graph embedding techniques, which, although able to integrate domain knowledge into models, often trade explainability for predictive performance, and examines recent advances in explainable approaches for semantic similarity, embeddings, and path-based reasoning.
LGOBE rOcid CEUR