New publication at ESWC 2026: SIDEKICK, a semantically integrated knowledge graph for drug safety
SIDEKICK, presented at ESWC 2026 by Mohammad Ashhad, Olga Mashkova, Ricardo Henao (Duke) and Robert Hoehndorf, builds an ontology-aligned knowledge graph of drug indications, contraindications and adverse reactions from 50,000+ FDA drug labels.
About
The Bio-Ontology Research Group presented "SIDEKICK: A Semantically Integrated Resource for Drug Effects, Indications, and Contraindications" at the 23rd Extended Semantic Web Conference (ESWC 2026). The paper, by Mohammad Ashhad, Olga Mashkova, Ricardo Henao (Duke University) and Robert Hoehndorf, introduces a new knowledge graph for pharmacovigilance and clinical decision support.
Existing drug-safety datasets typically depend on the MedDRA terminology, which limits semantic reasoning and interoperability with Semantic Web ontologies. SIDEKICK addresses this by standardising drug indications, contraindications and adverse reactions extracted from FDA Structured Product Labels, mapping the extracted terms to the Human Phenotype Ontology (HPO), the MONDO Disease Ontology, and RxNorm. The extraction workflow combines Large Language Model (LLM) extraction with Graph-Retrieval Augmented Generation (Graph RAG) for ontology mapping, and is applied across more than 50,000 drug labels.
The resulting graph outperforms the SIDER and ONSIDES databases when applied to drug repurposing by side-effect similarity. SIDEKICK is serialised as an RDF graph and uses the Semanticscience Integrated Ontology (SIO) as upper-level scaffolding to further improve interoperability, enabling automated safety surveillance and phenotype-based similarity analysis for drug repurposing.
The resource is available at sidekick.bio2vec.net (CC BY 4.0), archived on Zenodo (10.5281/zenodo.17779317), and published in the ESWC 2026 proceedings; DOI: 10.1007/978-3-032-25159-6_14.