Open Targets Platform

The Open Targets Platform is a publicly available knowledge base that systematically links potential drug targets (genes and proteins) to diseases, scoring each association from a wide range of integrated evidence. It draws together roughly two dozen independent public data sources covering common and rare genetic variation, somatic mutations, gene expression, pathways, animal models, literature text mining, and known drugs, then computes weighted association scores across hundreds of thousands of target-disease pairs. A recent release described around 865,000 unique target-disease pairs, giving a sense of the platform's scale.

You can access the data in several ways. There is a browsable web interface, a GraphQL API and a REST API for programmatic queries, a Google BigQuery instance for systematic analysis, and full bulk downloads distributed as Parquet (and JSON) files via EMBL-EBI FTP and mirrored on Google Cloud, Microsoft Azure Open Datasets, and the AWS Open Data Registry. Data is organised into structured entities such as targets, diseases, evidence, and associations.

For RAG and AI work, the association evidence and the rich entity annotations (target descriptions, disease definitions, tractability, safety, and drug data) are a strong grounding layer for biomedical and drug-discovery assistants. The structured Parquet tables suit building retrieval indexes and knowledge graphs, while the APIs support live lookups. Best treated as curated, structured evidence rather than free text: you will typically need to join and format entities into passages before chunking and embedding.

Watch-outs: the platform is versioned by periodic releases, so pin a release for reproducibility. The data is dedicated to the public domain under CC0, but some underlying source datasets carry their own terms, so check provenance if you redistribute derived subsets. For adjacent biomedical grounding, consider PubMed and PubTator, or ChEMBL for bioactivity data.