ChEMBL provider¶

ChEMBL contributes two kinds of Clinical Reports to Mira: curated drug indications and drug safety warnings. Both are read from the ChEMBL PostgreSQL database and already contain ChEMBL drug identifiers. Many indication and warning records also contain EFO disease or safety-outcome identifiers.

See ChEMBL 37 PostgreSQL database for the release, tables, selected columns, connection settings, and local data preparation.

What ChEMBL contributes¶

ChEMBL is the provider. The source of each report comes from the ChEMBL reference type, so reports supplied by ChEMBL can represent evidence originating from FDA, EMA, DailyMed, ATC, INN, USAN, ClinicalTrials.gov, or another curated reference.

ChEMBL contribution	Clinical Report `type`	Typical `origin`	Main content
Drug indications	`INDICATION`	Clinical trial, drug label, regulatory agency, or curated resource	A drug, disease, development stage, and supporting reference.
Drug warnings	`SAFETY`	Drug label, regulatory agency, or curated resource	A drug, warning or withdrawal status, jurisdiction, and optional mapped safety outcome.

Important

The provider/source distinction matters when filtering the output. To find every report obtained through ChEMBL, use provider == "ChEMBL"; to select an original evidence source, filter source instead.

Drug indications¶

ChEMBL indications join the drug_indication, molecule_dictionary, and indication_refs tables. Each retained reference becomes a traceable indication report with:

the preferred ChEMBL molecule name and identifier;
the ChEMBL EFO term and identifier;
the reference type, identifier, and URL;
the maximum phase reported by ChEMBL where the reference is not itself evidence of approval.

The reference type determines the report origin. ClinicalTrials.gov references become CLINICAL_TRIAL, DailyMed references become DRUG_LABEL, and FDA or EMA references become REGULATORY_AGENCY. Other reference types are represented as CURATED_RESOURCE.

Selection and stage rules¶

Rows without an efo_id are excluded.
Reference URLs beginning with www are excluded by the current adapter.
ATC, EMA, FDA, DailyMed, and PMDA references are treated as approval evidence.
INN and USAN references have an unknown clinical stage.
Other references use ChEMBL's max_phase_for_ind value, which Mira then maps to its shared clinical-stage categories.

Note

In the default Open Targets workflow, indication reports whose source is ClinicalTrials are removed before providers are combined. AACT is the selected provider for ClinicalTrials.gov study records in that workflow.

Drug warnings¶

ChEMBL warnings join the drug_warning, molecule_dictionary, and warning_refs tables. They become safety reports containing:

a mapped ChEMBL drug;
the warning type as phaseFromSource;
the warning year and countries, when supplied;
the supporting reference and URL;
an optional side effect or warning class represented with its EFO identifier when available.

DailyMed warnings have a DRUG_LABEL origin. FDA and EMA warnings have a REGULATORY_AGENCY origin. Other warning references use CURATED_RESOURCE.

Note

Warnings do not require a mapped side effect. Mira keeps a warning report when its drug and reference are present even if sideEffects is null.

Mapping behavior¶

ChEMBL supplies drugId directly from molecule_dictionary. Indication diseases also enter the pipeline with an EFO diseaseId. Warning side effects use the warning's EFO term and identifier where available, falling back to the warning class fields.

The shared entity-mapping stage preserves identifiers already supplied by the provider and attempts to fill remaining gaps across the combined report dataset.