ChEMBL

ChEMBL is a manually curated database of bioactive molecules with drug-like properties, maintained by the European Bioinformatics Institute (EMBL-EBI) at the Wellcome Genome Campus in Hinxton, UK. What distinguishes ChEMBL from a general chemical registry is that every entry carries a measured bioactivity value — a Ki, Kd, IC50, or EC50 — against a specific, named molecular target, sourced from medicinal-chemistry literature and deposited screening data, not just a chemical structure with no functional evidence attached. That distinction is the actual operational definition of the database: a compound can be extensively documented elsewhere and still not be meaningfully “in ChEMBL” until a curated activity measurement links it to a target.

Where ChEMBL came from

ChEMBL began as StARlite, a proprietary database built by the UK company Inpharmatica Ltd. EMBL-EBI acquired the underlying dataset in 2008 with Wellcome Trust funding and reorganized it into an open, publicly maintained resource, releasing it under the ChEMBL name in 2009. Early releases already ran to millions of bioassay measurements curated from tens of thousands of publications across a defined set of medicinal-chemistry journals — the scale that has made ChEMBL a standard reference point for drug-discovery bioinformatics rather than a niche academic dataset.

What a ChEMBL record actually contains

A ChEMBL entry links three things: a compound (with a standardized structure), a target (protein, cell line, or organism, tied to identifiers like UniProt accessions), and an assay result quantifying the interaction between them. Assay metadata records the experimental context — assay type, organism, cell line, publication source — so a researcher can filter results to, for example, human-only enzymatic assays rather than treating every reported activity value as directly comparable. This structure is what makes ChEMBL usable for quantitative structure-activity relationship (QSAR) modeling and machine-learning bioactivity prediction: the training signal is a real measured value tied to a defined target and assay condition, not an inferred or predicted one.

Access and licensing

ChEMBL data is released under a Creative Commons Attribution-ShareAlike license and is available through several access paths: the web interface for interactive browsing, bulk FTP/SQLite/MySQL/PostgreSQL downloads for full offline use, a REST web service for programmatic queries, and a SPARQL endpoint for semantic queries across the underlying RDF representation. Content is versioned by release number, and because target and assay coverage genuinely changes between releases, a data management plan or methods section citing ChEMBL should specify which release was used, the same way a software dependency would be version-pinned.

Where ChEMBL fits among named repositories

ChEMBL is a bioactivity database, not a general chemical registry — that role is filled by PubChem, which indexes ChEMBL as one of its many source databases alongside a far larger set of registered structures with no bioactivity requirement. A researcher citing “public chemistry data” without specifying which of the two they mean is eliding a real methodological distinction — PubChem answers “does this structure exist and what is known about it,” ChEMBL answers “what was measured about this compound’s activity against this target, and under what conditions.” A data management plan should name the specific resource and release version it depends on, the same accession-level specificity CASRAI’s own Data Management Plan guidance recommends for any named repository dependency.