Associations with Qualifiers
Chemical affects Gene Association
Here we take some real-world examples of knowledge statements from a variety of sources and describe how they would be represented in Biolink Model.
Examples
"Methionine deficiency results in increased expression of ADRB2 in adipose tissue"
In keeping with our modeling paradigms (Curating The Model), - Nodes should represent core domain concepts - Use qualifiers to compose full node semantics - The ‘core triple’ should remain true if qualifiers are ignored. (except 'negation') - Represent information consistently
We need to break this statement into its core components that are most likely going to be shared across many statements and data sources. In this example sentence, we can first break it apart into three parts or core components: - Methionine (biolink:subject) - affects (biolink:predicate) - ADRB2 (biolink:object)
This is our "core triple".
Some design decisions that went into selecting this core triple: - Use of the Biolink model does not require, always, the use of less specific subject and object concepts, some use knowledge graphs may choose to use a more specific subject and object category to fulfill their use cases. But for this example, we apply Biolink categories at a higher level in the core triple to increase the connectivity possibilities of the data.
- The "affects" predicate was chose here because we are being careful to use a predicate that keeps the core triple true despite our choice of subject and object. If we were to use "increases expression of" as a predicate here, we would be invalidating the "core tripe is true" principle (as our subject, "Methionine" is the opposite of "Methionine deficiency").
Despite our preference for connectivity, we don't want to lose the specificity of our original knowledge statement.
We can capture the nuance and specificity (expression, abundance) in qualifiers. The aspect of Methionine that we
want to capture is the "decreased abundance" of it. The aspect of ADRB2 that we want to capture is the "increased expression" of it.
These two aspects, "abundance" and "expression" are represented in the Biolink model as "aspect qualifiers".
The direction of the aspect, is represented in the Biolink model as "direction qualifiers". Finally, we can have a
qualifier that applies to the entirety of the fully qualified statement, an "anatomical context qualifier" that expresses
that this statement all takes place in the "adipose tissue."
Therefore, the fully qualified statement, using our core triple would be:
- in JSON format:
{
"id": "e0",
"category": "biolink:ChemicalAffectsGeneAssociation",
"subject": "CHEBI:16811", # methionine
"subject_aspect_qualifier": "abundance",
"subject_direction_qualifier": "decreased",
"predicate": "biolink:affects",
"qualified_predicate": "biolink:causes",
"object": "NCBIGene:154", # ADRB2
"object_aspect_qualifier" : "expression",
"object_direction_qualifier" : "increased",
"anatomical_context_qualifier": "UBERON:0001013" # adipose tissue,
"publications": ["PMID:123456"],
"original_knowledge_source": "infores:molepro"
}
Note: this is a representation of the fully qualified statement in an association. This association is not limited to subject, predicate, object and qualifiers. It can also have other slots, such as "original_knowledge_source" and "publications" that express the evidence and provenance of this statement.
More information on knowledge source retrieval provenance can be found in the Knowledge Source Retrieval guidelines.
- KGX flat file (edges.tsv)
subject, predicate, object, qualified_predicate, subject_aspect_qualifier, subject_direction_qualifier, object_aspect_qualifier, object_direction_qualifier, anatomical_context_qualifier
CHEBI:16811, biolink:affects, NCBIGene:154, biolink:causes, abundance, decreased, expression, increased, UBERON:0001013
- TRAPI (Translator API specification, knowledge graph representation)
https://github.com/NCATSTranslator/ReasonerAPI/blob/master/examples/Message/subject_and_object_qualifiers.json
{
"knowledge_graph": {
"nodes": {
"PUBCHEM.COMPOUND:6137": {
"categories": [
"biolink:ChemicalEntity"
],
"name": "Methionine"
},
"HGNC:286": {
"categories": [
"biolink:GeneOrGeneProduct"
],
"name": "ADRB2"
}
},
"edges": {
"x17770": {
"predicate": "biolink:affects",
"subject": "PUBCHEM.COMPOUND:6137",
"object": "HGNC:286",
"qualifiers": [
{
"qualifier_type_id": "biolink:subject_aspect_qualifier",
"qualifier_value": "abundance"
},
{
"qualifier_type_id": "biolink:subject_direction_qualifier",
"qualifier_value": "decreased"
},
{
"qualifier_type_id": "biolink:qualified_predicate",
"qualifier_value": "biolink:causes"
},
{
"qualifier_type_id": "biolink:object_aspect_qualifier",
"qualifier_value": "expression"
},
{
"qualifier_type_id": "biolink:object_direction_qualifier",
"qualifier_value": "increased"
}
]
}
}
}
}
"Fenofibrate binds to PPARA protein" - This is a simple Chemical interacts with Gene statement (no qualifiers needed)
{
"id": "e0",
"category": "biolink:ChemicalAffectsGeneAssociation",
"subject": "CHEBI:5001" # Fenofibrate
"predicate": "biolink:physically_interacts_with"
"object": "NCBIGene:5465" # PPARA
}
"Cyclophosphamide affects the hydroxylation of CYP2B6" - A simple chemical affects gene (aspect) Statement - no direction to the effect.
{
"id": "e0",
"category": "biolink:ChemicalAffectsGeneAssociation",
"subject": "CHEBI:16811" # Cyclophosphamide,
"predicate": "biolink:affects",
"object": "NCBIGene:154" # CYP2B6,
"object_aspect_qualifier": "hydroxylation"
}
"Bisphenol A results in decreased degradation of ESR1 protein" - A Statement where the effect has a direction (decreased)
{
"id": "e0",
"category": "biolink:ChemicalAffectsGeneAssociation",
"subject": "CHEBI:16811" # Bisphenol A,
"predicate": "biolink:affects",
"qualified_predicate": "biolink:causes",
"object": "NCBIGene:2099" # ESR1,
"object_aspect_qualifier": "degradation",
"object_direction_qualifier": "decreased"
}
"Bisphenol A is associated with decreased degradation of ESR1 protein" - A (hypothetical) chemical associated_with gene (aspect) Statement with same S/O concepts as above
{
"id": "e0",
"category": "biolink:ChemicalAffectsGeneAssociation",
"subject": "CHEBI:16811" # Bisphenol A,
"predicate": "biolink:associated_with",
"object": "NCBIGene:2099" # ESR1,
"object_aspect_qualifier": "degradation",
"object_direction_qualifier": "decreased"
}
"Progesterone metabolites cause decreased methylation of APP promoter mutant forms" - A more complex example where a metabolite of the specified chemical is the effector of a heavily qualified Statement object.
{
"id": "e0",
"category": "biolink:ChemicalAffectsGeneAssociation",
"subject": "CHEBI:17026" # Progesterone,
"predicate": "biolink:affects",
"qualified_predicate": "biolink:causes",
"object": "NCBIGene:351" # APP,
"object_aspect_qualifier": "methylation",
"object_direction_qualifier": "decreased",
"object_part_qualifier": "promoter",
"object_form_or_variant_qualifier": "mutant"
}
“Hexachlorobenzene analog causes increased methylation of CDKN2A enhancer alternative form” - Another complex example where an analog of a specified chemical is the effector of a heavily qualified Statement object.
{
"id": "e0",
"category": "biolink:ChemicalAffectsGeneAssociation",
"subject": "CHEBI:16811" # Hexachlorobenzene,
"subject_form_or_variant_qualifier": "analog",
"predicate": "biolink:affects",
"qualified_predicate": "biolink:causes",
"object": "NCBIGene:1029" # CDKN2A,
"object_aspect_qualifier": "methylation",
"object_direction_qualifier": "increased",
"object_part_qualifier": "enhancer",
"object_form_or_variant_qualifier": "modified"
}
"Fenofibrate is an agonist of PPARA protein” - Chemical increases' gene activity, via a specific control mechanism (agonist)
{
"id": "e0",
"category": "biolink:ChemicalAffectsGeneAssociation",
"subject": "CHEBI:5001" # Fenofibrate,
"predicate": "biolink:affects",
"qualified_predicate": "biolink:causes",
"object": "NCBIGene:5465" # PPARA,
"object_aspect_qualifier": "activity",
"object_direction_qualifier": "increased",
"causal_mechanism_qualifier": "agonism"
}