GeneCore

GeneCore provides access to 43K+ harmonized gene records integrated from HGNC, NCBI, UniProt, and Ensembl. Each record carries a rich set of external identifiers: Ensembl, NCBI Entrez, UniProt, MANE Select transcripts, OMIM, Orphanet, and IUPHAR, plus the HGNC gene family it belongs to.

For drug-target genes, records are enriched with comprehensive target intelligence spanning four key domains (see Target intelligence detail):

• Clinical-Precedent Tractability: Druggability and clinical-readiness buckets across small molecules, antibodies, and other modalities.
• Curated Target Safety: Known target-safety signals, clinical adverse events, and known target-modulation effects.
• Genetic & Evolutionary Constraint: Population-level intolerance to loss of function, synonymous/missense mutations, and constraint scores.
• Cellular Essentiality: Gene dependency profiles and essentiality mapping across disease model cell lines.

These records also feature an opt-in Swiss-Prot protein-annotation block and are cross-linked to their targeting drugs (DrugCore).

Endpoints

Search — GET /v1/cores/genecore/records

Keyword search across gene symbols, names, synonyms, gene families, RefSeq functional summaries, UniProt keywords, and ChEMBL target class, with optional filters for gene type and target intelligence (druggability, tractability, target class, safety, genetic constraint). Beyond exact symbols/names, free-text concept queries (e.g. tyrosine kinase, GPCR, apoptosis) match the controlled-vocabulary keyword and class fields.

curl "https://api.amass.tech/api/v1/cores/genecore/records?query=GLP1R&limit=5" \
  -H "Authorization: Bearer amass_YOUR_KEY"

Parameters:

Example with filters — protein-coding genes matching a keyword:

curl "https://api.amass.tech/api/v1/cores/genecore/records\
?query=glucagon+receptor\
&geneType=PROTEIN_CODING\
&limit=20" \
  -H "Authorization: Bearer amass_YOUR_KEY"

Repeat geneType to match ANY of several values, e.g. geneType=PROTEIN_CODING&geneType=NCRNA.

Example with target-intelligence filters — druggable enzymes with an approved small-molecule precedent:

curl "https://api.amass.tech/api/v1/cores/genecore/records\
?query=kinase\
&targetClass=ENZYME\
&tractabilityModality=SMALL_MOLECULE\
&tractabilityStage=APPROVED_DRUG\
&isDruggable=true\
&limit=20" \
  -H "Authorization: Bearer amass_YOUR_KEY"

Response:

{ "data": [ { "amassId": "AMGC_...", "symbol": "GLP1R", "..." }, ... ] }

Get by ID — GET /v1/cores/genecore/records/{amassId}

Fetch a single gene record by its Amass ID.

curl "https://api.amass.tech/api/v1/cores/genecore/records/AMGC_abc123\
?include=referencesDrugCore" \
  -H "Authorization: Bearer amass_YOUR_KEY"

Response:

{ "data": { "amassId": "AMGC_...", "symbol": "...", "..." } }

Returns 404 if not found, or 400 if the Amass ID is malformed.

Lookup — POST /v1/cores/genecore/records/lookup

Convert public gene identifiers to Amass IDs in batch. Each item carries exactly one identifier key — an empty item, or one with several identifier keys, is rejected.

curl -X POST "https://api.amass.tech/api/v1/cores/genecore/records/lookup" \
  -H "Authorization: Bearer amass_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "items": [
      { "ensemblGeneId": "ENSG00000146648" },
      { "symbol": "EGFR" },
      { "uniprotId": "P00533" },
      { "ensemblGeneId": "ENSG99999999999" }
    ]
  }'

Response:

{
  "data": [
    {
      "input": { "ensemblGeneId": "ENSG00000146648" },
      "amassIds": ["AMGC_abc123"]
    },
    {
      "input": { "symbol": "EGFR" },
      "amassIds": ["AMGC_abc123"]
    },
    {
      "input": { "uniprotId": "P00533" },
      "amassIds": ["AMGC_abc123"]
    },
    {
      "input": { "ensemblGeneId": "ENSG99999999999" },
      "error": { "code": "NOT_FOUND", "message": "Gene record not found." }
    }
  ]
}

Identifiers are matched against the genes namespace before an Amass ID is issued, so unknown IDs come back as NOT_FOUND rather than a dangling AMGC_. Most identifiers resolve to a single gene; a shared uniprotId/omimId may return more than one AMGC_. Individual items can fail without failing the whole request — always check each item for an error field.

Record Schema

Default fields (always returned)

The remaining default fields are target-intelligence objects — returned by default, but null when no data for the gene (see Target intelligence detail for each object's shape):

Optional fields

Request these with the include parameter. Repeat for multiple: ?include=protein&include=referencesDrugCore

Reference field semantics

╔══════════════ GeneCore (genes) ══════════════╗
║                                              ║
║                ┌──────────┐                  ║
║                │  AMGC_X  │                  ║
║                └──────────┘                  ║
║                                              ║
╚══════════════════════════════════════════════╝
                       │
               referencesDrugCore
               = [AMDC_d01, AMDC_d02]
                       │
                       ▼
        ╔═══════ DrugCore (drugs) ═══════╗
        ║                                ║
        ║      AMDC_d01      AMDC_d02     ║
        ║                                ║
        ╚════════════════════════════════╝

• Cross-core arrows leave GeneCore — referencesDrugCore targets start with AMDC_.
• Inside DrugCore, the linked records carry their own reference fields (trials, literature, regulatory authorizations) — see DrugCore.
• GeneCore records have no intra-core hierarchy; every link points out to another core.

Target intelligence detail

Each record carries target-intelligence objects, returned by default and null when no target intelligence is recorded for the gene. The objects below describe the returned shape; each also documents its matching search filter. Many genes (especially non-protein-coding) have no target intelligence at all.

Tractability

tractability is keyed by drug modality — smallMolecule, antibody, protac, otherClinical — and each modality splits its satisfied druggability buckets into two lanes:

Filter — two enum params, each repeatable (OR), combined as a cross-product where an omitted dimension means "any":

• tractabilityModality: SMALL_MOLECULE, ANTIBODY, PROTAC, OTHER_CLINICAL
• tractabilityStage: APPROVED_DRUG, ADVANCED_CLINICAL, PHASE_1_CLINICAL

tractabilityModality=SMALL_MOLECULE alone matches any clinical stage; pair with tractabilityStage to require a stage. Only the three clinical-precedent stages are filterable — the predictive lane is returned but not filterable. isDruggable=true is a shortcut for "any satisfied small-molecule or antibody bucket".

Target class

targetClass is the ChEMBL protein-classification hierarchy collapsed to an ordered path:

Filter — targetClass, a repeatable (OR) enum of the top-level (broadest) classes only; the full path is still returned, but deeper levels (e.g. Kinase) are not filterable. Values: ENZYME, MEMBRANE_RECEPTOR, ION_CHANNEL, TRANSPORTER, TRANSCRIPTION_FACTOR, EPIGENETIC_REGULATOR, SECRETED_PROTEIN, SURFACE_ANTIGEN, STRUCTURAL_PROTEIN, ADHESION, OTHER_CYTOSOLIC_PROTEIN, OTHER_NUCLEAR_PROTEIN, AUXILIARY_TRANSPORT_PROTEIN, UNCLASSIFIED_PROTEIN.

Safety liabilities

safetyLiabilities is an array of curated target-safety signals, each:

Filter — hasSafetyLiabilities=true keeps only genes with at least one liability. The event vocabulary is curated free text, so it is not filterable by value — filter on presence and read the events from the response.

Genetic constraint (gnomAD v4.0)

gnomadConstraint summarizes gnomAD v4.0 Gene Constraint, keyed by variant class — synonymous, missense, and lossOfFunction. Each class carries observed vs. expected variant counts and their ratio (oe, with oeLower / oeUpper bounds; the synonymous / missense classes also carry a constraint Z-score under constraintZ). The headline loss-of-function metrics live under lossOfFunction:

LOEUF guidance. LOEUF is a continuous metric of a gene's intolerance to loss-of-function variation, and we encourage using it as such. However, if your application requires a binary threshold to define loss-of-function (LoF) constrained genes, use these recommended cutoffs for gnomAD v4.0 (the version displayed in this API):

• LOEUF: Use LOEUF < 0.6 (note that the distribution shifted from < 0.35 in gnomAD v2.1.1).
• pLI: Use pLI ≥ 0.9 (remains stable across versions).
• LOEUF Deciles: Alternatively, filter by the first decile, as relative rankings are highly consistent across releases.

Filter — maxConstraintLoeuf keeps genes whose LOEUF (lossOfFunction.loeuf) is ≤ the value you pass (a lower bound selects more-constrained genes):

curl "https://api.amass.tech/api/v1/cores/genecore/records\
?query=kinase\
&maxConstraintLoeuf=0.6\
&limit=20" \
  -H "Authorization: Bearer amass_YOUR_KEY"

Essentiality

depmapEssentiality summarizes the CRISPR dependency screens for the gene:

A more negative gene-effect means a stronger dependency. Filter — isEssential=true keeps genes flagged as a dependency.

Protein (include=protein)

Opt in with ?include=protein. protein is one representative UniProt Swiss-Prot entry per gene (prose fields from the canonical entry; list fields unioned across matched entries), grouped into four blocks:

protein is null when the gene encodes no reviewed Swiss-Prot protein.

Filter Values

Gene Type

NCBI gene type / biotype (joined from NCBI Gene by Entrez ID). Use with the geneType search filter or read from the geneType field.

See Also

• DrugCore — Drug and molecule records
• TrialCore — Clinical trial records
• BiomedCore — Biomedical literature records
• RegulatoryCore — FDA/EMA regulatory authorization records
• API Workflows — See GeneCore in action: target prioritization, genetic constraint filters, and mapping targets to clinical pharmacology
• Overview — Errors and rate limits
• Quickstart — Getting started with examples