PharmKG is a multi-relational, attributed biomedical knowledge graph benchmark that integrates curated databases and literature-derived relationships to link genes, drugs (chemicals), and diseases. It contains ~500,000 typed relations across ~7,600–8,000 disambiguated entities (29 relation types), with domain-specific features for entities from multi-omics sources (e.g., gene expression, chemical structure, disease embeddings). It provides baselines for knowledge graph embedding and a neural network-based method, and supports downstream tasks such as drug repurposing and target identification.
Graph at a glance
Products & downloads
No node-type (Biolink category) breakdown is recorded for this graph. Total nodes: 7,603.
No edge-type (predicate) breakdown is recorded for this graph. Total edges: 500,958.