Drug-Repurposing Hub Reference Ingest Guide¶
Source Information¶
InfoRes ID: infores:drug-repurposing-hub
Description: The Drug Repurposing Hub is a curated and annotated collection of FDA-approved drugs, clinical trial drugs, and selected pre-clinical tool compounds, produced by the Broad Institute. Each compound is annotated with its clinical development status, mechanism of action, known protein target(s), disease area, and disease indication(s). Because the compounds are annotated and well characterized, the collection serves as a tool to explore known biological targets and pathways, and a way to discover new biological insights, better understand disease mechanisms, and identify new activities of existing molecules.
Citations: - Corsello SM, Bittker JA, Liu Z, Gould J, McCarren P, Hirschman JE, Johnston SE, Vrcic A, Wong B, Khan M, Asiedu J, Narayan R, Mader CC, Subramanian A, Golub TR. The Drug Repurposing Hub: a next-generation drug library and information resource. Nature Medicine. 23, 405-408 (2017). https://doi.org/10.1038/nm.4306
Data Access Locations: - https://repo-hub.broadinstitute.org/repurposing#download-data
Data Provision Mechanisms: file_download
Data Formats: tsv
Data Versioning and Releases: Releases are irregular and not on a fixed cadence. Latest version: 8/19/2025. Archived versions: 3/24/2020, 9/7/2018, 5/16/2018, 3/27/2017.
Additional Notes: ['The ingest currently pins the 20200324 drug-annotation file and the 20240610 sample-annotation file (see download.yaml). A future iteration should move to the latest 2025-08-19 release.']
Ingest Information¶
Ingest Categories: primary_knowledge_provider
Utility: Provides manually curated compound-to-target, compound-to-disease-indication, and compound-to-chemical-role annotations for a well-characterized library of approved, clinical-stage, and pre-clinical compounds. This supports Translator drug repurposing, mechanism-of-action, and treatment-prediction use cases.
Scope: FDA-approved drugs, clinical trial drugs, and selected pre-clinical tool compounds, annotated with mechanism of action, protein target(s), clinical phase, disease area, and disease indication(s).
Relevant Files¶
| File Name | Location | Description |
|---|---|---|
| repo-drug-annotation.txt | http://repo-hub.broadinstitute.org/public/data/repo-drug-annotation-20200324.txt | One row per compound, providing curated annotations including clinical phase, mechanism of action, protein target(s), disease area, and disease indication(s). Source of all edges in this ingest. |
| repo-sample-annotation.txt | http://repo-hub.broadinstitute.org/public/data/repo-sample-annotation-20240610.txt | One row per compound sample, providing compound identity information (PubChem CID, InChIKey, vendor name). Used to assign identifiers to the chemical (subject) nodes that the drug-annotation edges attach to. |
Included Content¶
| File Name | Included Records | Fields Used |
|---|---|---|
| repo-drug-annotation.txt | All compound rows whose pert_iname matches a sample loaded from repo-sample-annotation.txt and that yield at least one target or indication association. | pert_iname, clinical_phase, moa, target, disease_area, indication |
| repo-sample-annotation.txt | All sample rows that resolve to a chemical identifier (a numeric PubChem CID, or a valid InChIKey when no CID is present). | pert_iname, InChIKey, vendor_name, pubchem_cid |
Filtered Content¶
| File Name | Filtered Records | Rationale |
|---|---|---|
| repo-sample-annotation.txt | Rows with neither a numeric PubChem CID nor a valid InChIKey (failing the InChIKey format regex). | These samples cannot be assigned a resolvable chemical identifier, so no node can be created. |
| repo-drug-annotation.txt | Rows whose pert_iname does not match a loaded sample, and rows that produce no target or indication edge. | Without a matching identified compound, or without at least one mappable target/indication, no edge can be created. Indications not present in indications_config.json and target gene symbols not present in target_config.json (unmappable to HGNC) are skipped. |
Future Content Considerations¶
node_property_content: The 'disease_area' field is read from the source but not currently emitted as a node property (the assignment is commented out as a TODO in create_disease_association / the chemical node build). Consider adding it as a node property once a suitable Biolink slot (or a registered attribute) is available. - Relevant files: repo-drug-annotation.txt
edge_property_content: clinical_approval_status and max_research_phase are computed via clinical_approval_map and research_phase_map but are currently commented out as TODOs; wire them onto the treatment edges as edge properties in a future iteration. - Relevant files: repo-drug-annotation.txt
Target Information¶
Edge Types¶
| Subject Categories | Predicate | Object Categories | Knowledge Level | Agent Type | UI Explanation |
|---|---|---|---|---|---|
| biolink:ChemicalEntity, biolink:SmallMolecule, biolink:MolecularMixture | biolink:Gene | knowledge_assertion | manual_agent | The Drug Repurposing Hub provides manually curated annotations of the known protein target(s) and mechanism of action for each compound in its library. This edge was created from a compound's curated 'target' gene-symbol annotation (mapped to an HGNC gene), with the compound's curated mechanism-of-action ('moa') text captured on the edge as its description. Because the target is a curated assertion rather than a statistical or computational inference, it is represented using the Biolink 'affects' predicate with a knowledge_assertion knowledge level and manual_agent agent type. | |
| biolink:ChemicalEntity, biolink:SmallMolecule, biolink:MolecularMixture | biolink:DiseaseOrPhenotypicFeature | knowledge_assertion | manual_agent | The Drug Repurposing Hub provides manually curated annotations of the disease indication(s) for each compound, along with the compound's highest clinical development phase. This edge was created from a compound's curated 'indication' annotation, mapped to a disease/phenotype term. The specific Biolink predicate is selected from the curated feature/action of the annotation combined with the compound's clinical_phase: launched drugs use 'treats'; compounds in clinical or pre-clinical study use 'in_clinical_trials_for' or 'in_preclinical_trials_for'; withdrawn or unspecified-phase compounds use the broad 'treats_or_applied_or_studied_to_treat'; and annotations curated as ameliorative or diagnostic use 'ameliorates_condition' or 'diagnoses' respectively. All are curated assertions (knowledge_assertion / manual_agent). | |
| biolink:ChemicalEntity, biolink:SmallMolecule, biolink:MolecularMixture | biolink:ChemicalEntity | knowledge_assertion | manual_agent | The Drug Repurposing Hub annotates some compounds with a functional role or drug category (e.g. 'antiseptic', 'contrast agent', 'antiviral') rather than a specific disease indication. When a compound's curated 'indication' annotation denotes such a role, it is represented as a 'has_chemical_role' edge linking the compound to a chemical-role concept (typically a ChEBI or NCIT term). This is a curated assertion (knowledge_assertion / manual_agent). |
Node Types¶
| Node Category | Source Identifier Types | Additional Notes |
|---|---|---|
| biolink:ChemicalEntity | PUBCHEM.COMPOUND, INCHIKEY, CHEBI, NCIT | Compound (subject) nodes are identified by PubChem CID (PUBCHEM.COMPOUND) or InChIKey, and carry name, synonym (from the sample vendor name when distinct), and xref (InChIKey) properties. Chemical-role (object) nodes of has_chemical_role edges are also created as biolink:ChemicalEntity, identified by ChEBI or NCIT, and carry a name. Downstream normalization may reassign some compound nodes to biolink:SmallMolecule or biolink:MolecularMixture. |
| biolink:DiseaseOrPhenotypicFeature | MONDO, HP, NCIT, DOID, EFO, NCBITaxon | Disease/indication identifiers are heterogeneous; the most common prefixes are MONDO and HP, followed by NCIT, with a long tail (DOID, EFO, NCBITaxon, OMIT, SNOMED, Orphanet, and others). |
| biolink:Gene | HGNC |
Future Modeling Considerations¶
qualifiers: Map mechanism-of-action description text to qualifiers on the compound-affects-gene edges (e.g. 'serine protease inhibitor' => causal_mechanism_qualifier: inhibition).
predicates: Decompose the compound-to-disease-indication edge type into separate EdgeType objects, since its predicates (treats, treats_or_applied_or_studied_to_treat, in_clinical_trials_for, in_preclinical_trials_for, ameliorates_condition, diagnoses) are not a single predicate-plus-descendants hierarchy as required by the schema for a shared EdgeType.
edge_properties: Emit the currently-computed-but-commented-out clinical_approval_status and max_research_phase as edge properties on treatment edges, and consider a dedicated mechanism_of_action property (registered in the attributes catalog with a biolink prefix) in place of the generic biolink:description slot for compound-affects-gene edges.
Provenance Information¶
Contributors: - Vlado Dancik: code author, domain expertise - Kevin Schaper: code support - Evan Morris: code support - Sierra Moxon: data modeling, code support - Matthew Brush: data modeling, domain expertise
Artifacts: - RIG review sub-issue: https://github.com/NCATSTranslator/translator-ingests/issues/419 - RIG review epic: https://github.com/NCATSTranslator/translator-ingests/issues/407