Biolink Model

BioLink, Datamodel, NCATS, NCATS-Translator, gene-ontology, json-api, monarchinitiative, owl, specification, standard, yaml
pip install biolink-model==1.2.5


Python 3.7 Build Status DOI Join the chat at


BioLink Model

Quickstart docs:

See Introduction to the BioLink datamodel slides for a background on the BioLink Model and its inception.

Conversion/validation code:


The purpose of the BioLink model is to provide a high level datamodel of biological entities (genes, diseases, phenotypes, pathways, individuals, substances, etc), their properties, relationships, and enumerate ways in which they can be associated.

The representation is independent of storage technology or metamodel (Solr documents, neo4j/property graphs, RDF/OWL, JSON, CSVs, etc). Different mappings to each of these are provided.

The specification of the reference BioLink model is a single YAML file following a custom meta-model. The basic elements of the YAML are:

  • definitions of upper level classes representing both 'named thing' (genes, phenotypes, etc) and 'association'
  • definitions of slots (aka properties) that can be used to relate members of these classes to other classes or data types

This datamodel is being used in the NCATS Translator project. Not all these elements in the datamodel are used by the Translator; only a subset.

Entity (Node) Types

Protege view: img

Association Hierarchy


We divide these into relationship types (which connect two nodes together), node properties and edge properties.


See biolink json-ld context to find out more about identifier prefixes and their mappings.

Mapping to specific database and modeling platforms

Neo4J Mapping

Refer to mapping to neo4j for strategies on representing BioLink model in a Neo4j database.

RDF Mapping

Refer to mapping to RDF for strategies on representing BioLink model in a RDF triple store.


The datamodel source is biolink-model.yaml. This is a YAML file that is intended to be relatively simple to view and edit in its native form.

The yaml definition is currently used to derive:

We leverage existing frameworks where possible. E.g json-schema allows codegen to other languages.

Additionally, this repo contains the metamodel definition of itself in YAML, together with code for working with datamodels. In theory this could be used in other domains but there is no plan for this at the moment.


See metamodel for details of the metamodel.

Make and build instructions

Prerequisites: Python 3.7+

To install,

make install

If you make changes to biolink-model.yaml then be sure to run the Makefile to generate up-to-date artifacts and documentation.


Note: the Makefile requires jsonschema2pojo.

If you are on a Mac, it can be installed using brew:

brew install jsonschema2pojo

Usage in existing projects

Case study: gene expression in Monarch

Currently this is documented in the ingest artifacts repo, using non-computable cmap images:

bgee model

And also by the gene-anatomy cypher query which maps graphs conforming to the pattern to denormalized tuples for indexing in Solr

In the BioLink model this is explicitly represented using the gene to expression site association class definition in the model

  - name: gene to expression site association
    is_a: association
    description: >-
      An association between a gene and an expression site, possibly qualified by stage/timing info
    see_also: ""
      - slot: subject
        type: gene or gene product
        description: "gene in which variation is correlated with the phenotypic feature"
      - slot: object
        type: anatomical entity
        description: "location in which the gene is expressed"
        subclass_of: UBERON:0001062
          - value: UBERON:0002037
            description: cerebellum
      - slot: relation
        description: "expression relationship"
        subproperty_of: "RO:0002206"
      - slot: stage
        type: developmental stage
        description: "stage at which the gene is expressed in the site"
          - value: UBERON:0000069
            description: larval stage
      - slot: quantifier
        description: >-
          can be used to indicate magnitude, or also ranking

This is used to generate various artifacts such as

  • golr view definition
    • (which is itself later compiled to Solr XML using the BBOP-GOlrframework)
  • java class
    • generated from json-schema, so inheritance is unfolded
    • in future we may generate directly

Auto-generated UML diagram:


Auto-generated GraphQL definition:

type GeneToExpressionSiteAssociation {
  qualifiers: [String]
  stageQualifier: LifeStage
  objectExtensions: [PropertyValuePair]
  hasEvidence: String
  publications: [Publication]
  object: AnatomicalEntity!
  hasEvidenceType: EvidenceType
  hasEvidenceGraph: String
  providedBy: Provider
  label: String
  relation: String!
  negated: String
  subject: GeneOrGeneProduct!
  id: String!
  quantifierQualifier: String
  associationType: String
  subjectExtensions: [PropertyValuePair]

Auto-generated JSON Schema snippet:

        "GeneToExpressionSiteAssociation": {
            "description": "An association between a gene and an expression site, possibly qualified by stage/timing info. TBD: introduce subclasses for distinction between wild-type and experimental conditions?",
            "properties": {
                "association_type": {
                    "description": "connects an association to the type of association (e.g. gene to phenotype)",
                    "type": "string"
                "has_evidence": {
                    "description": "connects an association to an instance of supporting evidence",
                    "type": "string"
                "has_evidence_graph": {
                    "description": "connects an association to a graph object including a path from subject to object",
                    "type": "string"
                "has_evidence_type": {
                    "description": "connects an association to the class of evidence used",
                    "type": "string"
                "id": {
                    "type": "string"
                "label": {
                    "description": "A human-readable name for a thing",
                    "type": "string"
                "negated": {
                    "description": "if set to true, then the association is negated i.e. is not true",
                    "type": "string"
                "object": {
                    "description": "connects an association to the object of the association. For example, in a gene-to-phenotype association, the gene is subject and phenotype is object.",
                    "type": "string"
                "object_extensions": {
                    "description": "Additional relationships that are true of the object in the context of the association. For example, if the object is an anatomical term in an expression association, the object extensions may include part-of links",
                    "items": {
                        "type": "string"
                    "type": "array"
                "provided_by": {
                    "description": "connects an association to the agent (person, organization or group) that provided it",
                    "type": "string"
                "publications": {
                    "description": "connects an association to publications supporting the association",
                    "items": {
                        "type": "string"
                    "type": "array"
                "qualifiers": {
                    "description": "connects an association to qualifiers that modify or qualify the meaning of that association",
                    "items": {
                        "type": "string"
                    "type": "array"
                "quantifier_qualifier": {
                    "description": "A measurable quantity for the object of the association",
                    "type": "string"
                "relation": {
                    "description": "the relationship type by which a subject is connected to an object in an association",
                    "type": "string"
                "stage_qualifier": {
                    "description": "stage at which expression takes place",
                    "type": "string"
                "subject": {
                    "description": "connects an association to the subject of the association. For example, in a gene-to-phenotype association, the gene is subject and phenotype is object.",
                    "type": "string"
                "subject_extensions": {
                    "description": "Additional relationships that are true of the subject in the context of the association. For example, if the subject is a gene product in a functional association, the subject extensions may represent  an isoform or a specific post-translational state",
                    "items": {
                        "type": "string"
                    "type": "array"
            "required": [],
            "title": "GeneToExpressionSiteAssociation",
            "type": "object"


Why not use X as the modeling framework?

Why invent our own yaml and not use JSON-Schema, SQL, UML, ProtoBuf, OWL, etc.

Each of these is tied to a particular formalisms, e.g. JSON Schema to trees, OWL to open world logic. There are various impedance mismatches in converting between these. The goal was to develop something simple and more general that is not tied to any one serialization format or set of assumptions.

There are other projects with similar goals, e.g Schema Salad

It may be possible to align with these.

Why not use X as the datamodel

Here X may be BioSchemas, some upper ontology (BioTop), UMLS metathesaurus, bio*, various other attempts to model all of biology in an object model.

Currently as far as we know there is no existing reference datamodel that is flexible enough to be used here.