RDF data structure

WARNING - new RDF schema has been deployed and documentation is being updated at the moment.

The MolMeDB RDF dataset is divided into four domains according to their main object of focus - substances (split between substances themselves and substance attributes), membrane interactions, transporter interactions and references (thoroughly described below). Each domain has its own namespace of IRIs with a common prefix defined below in Table 1. Namespace values are used as shorthand for prefixes, as described in the previous chapter.

Table 1. Domains of MolMeDB RDF dataset.

Resources belonging to these domains are described using RDF triples where the IRI of the resource is the subject, the type of property is described by predicate, and its value represents the object. For example, the fact that molecule MM00040 has a name Caffeine is expressed as

mmdbmol:MM00040 rdfs:label “Caffeine”^^xsd:string - shown in Ontological Background - Figure 1.

All the described resources are typed, which means they have a specific class determined by rdf:type predicate e.g. caffeine is part of the dataset screened entity determined by the term from BioAssay ontology bao:BAO_0000076. In the final dataset, this relation is expressed as

mmdbmol:MM00040 rdf:type bao:BAO_0000076

Values of properties (objects) can be divided into four categories:

• literal value - string in quotation marks. If typed (language string, number, date etc.), the type is expressed by term after "^^". See the example above.
• typed term from MolMeDB RDF - value is an IRI of another resource from the dataset and has its own description. For example
  mmdbmol:MM00040 has molecular weight mmdbsub:MM00040_molecular_weight, which itself is further described by another triple in the dataset.
• term from ontology - value is an IRI of a resource not further described in the dataset but described in ontologies. Commonly used for units or types. For example, mmdbmol:MM00040 is of screened entity type, or mmdbmol:MM00040 rdf:type bao:BAO_0000076.
• term from another RDF database - value is an IRI of a resource not further described in the dataset but described in another RDF database. Commonly used to link resources in one dataset with another. For example, in MolMeDB RDF, Caffeine molecule record has IRI mmdbmol:MM00040 and in PubChem RDF exists corresponding IRI pubchem:CID2519. The fact that both represent the same resource is expressed by the triple mmdbmol:MM00040 skos:exactMatch pubchem:CID2519.

How to read the subdomain description?

Each domain consists of one or more resource types. For example, Substances domain consists of substances themselves and attributes of these substances. The resource of each type can be described as triples, which have the resource as a subject and can have predicates and objects particular to each resource type.

This document provides a description of each resource type belonging to the subdomains and lists of predicates and types of objects which can appear in triples describing them. For quick reference, tables for each resource type are provided.

Table 2. Explanation of reference tables for resource description.

Tables should be read in a following manner:

• A resource of the type can be described by any number of triples using the predicate listed in the "Predicate" column.
• Each triple can have an object particular to the predicate described by the "Object type" column.
• Object type categories are colour coded (see above).
• The meaning of each object type is explained in the "Description" column.
• The example column shows one example triple describing a resource of the type using the predicate on the same row.

For example, let's see part of a table for substance description (Table 3).

Table 3. Part of a reference table for substance description.

Table 3 shows that any substance can be described by any number of triples with the substance as a subject and sio:SIO_000008 (has attribute) as a predicate. Object types are blue coloured, which means that objects themselves are resources described in MolMeDB, and they can have either sio:CHEMINF_000018 (SMILES) or sio:CHEMINF_000059 (InChiKey) type. The example shows a triple saying that caffeine has a SMILES attribute.

Substances

Substances domain is split between two namespaces. First, mmdbmol: with https://identifiers.org/molmedb/ prefix, is used for IRIs of substances themselves and the second, mmdbsub: with https://rdf.molmedb.upol.cz/substance/ prefix, is used for IRIs of substance attributes. It is planned to allow for the use of mmdbsub: namespace for substances in the near future.

Example

Caffeine has IRI https://identifiers.org/molmedb/MM00040 (mmdbmol:MM00040 in compact form). A molecular weight of caffeine has IRI https://rdf.molmedb.upol.cz/substance/MM00040_molecular_weight (mmdbsub:MM00040_molecular_weight). Right now, https://rdf.molmedb.upol.cz/substance/MM00040 (mmdbsub:MM00040) is not a valid IRI in MolMeDB RDF, but it is planned to be equivalent to mmdbmol:MM00040 in the near future.

All substances are of type "screened entity" (bao:BAO_0000076). Possible matches to terms in other RDF databases (PubChem, ChEMBL, ChEBI and wwPDB) are expressed by triples with the "exact match" (skos:exactMatch) predicate. Other attributes of molecules are typed terms in MolMeDB RDF and are linked by the term "has attribute" (sio:SIO_000008). An attribute of a substance is described by its own set of triples, where the attribute acts as a subject.

A substance record can become obsolete and can be replaced by a new one. MolMeDB RDF keeps obsolete IRIs functional. Obsolete substance record has stated obsolescence reason  (obo:IAO_0000231) as "terms merged" (obo:IAO_0000227) and is linked to an up-to-date record using "term replaced by" (obo:IAO_0100001) predicate.

Table 4. Types of links describing molecules.

Substance attributes

Attributes are described by triples asserting their type and value (sio:SIO_000300). Possible types of attributes are listed as part of Table 5. Value can be a literal object of string or numeric (float) type. Molecular weight also has a unit (sio:SIO_000221) with the dalton (obo:NCIT_C41127) value.

Table 5. Types of links describing molecular attributes.

Membrane Interactions

Membrane interactions use mmdbint: namespace and are of "pharmacokinetic assay" (bao:BAO_0002182) type. Their description is based mainly on BioAssay Ontology (BAO), which uses a system where each bioassay is linked to one or more measure groups allowing for the multiplexing of different screened molecules on the same target. Measure groups are then linked to the final results, which are described by endpoints. In the case of MolMeDB RDF, all measure groups are considered independent bioassays also - any bioassay is linked precisely to one measure group (type bao:BAO_0000040) by "has measuregroup" (bao:BAO_0000209) predicate.

Relationship between a screened molecule (type bao:BAO_0000076) and relevant membrane model (type mmdbvoc:MembrameModel) is provided by triple with "has participant" (bao:BAO_0090012) predicate. Both molecule and membrane are further described in the dataset. Membranes have set more specific membrane model classes from MolMeDB vocabulary with more general classes up to mmdbvoc:MembraneModel explicitly expressed for easier querying.

The data source is expressed using cito:citesAsDataSource predicate, which links to either a publication (typedcterms:BibliographicResource) or a database (dcmitypes:Dataset); both are further described in the dataset.

Table 6. Types of links describing molecule-membrane interaction bioassays.

Membrane interaction measure groups

Links from the measure group (type bao:BAO_0000040) provide a detailed description of the measurement and they are linked to parental bioassay (type bao:BAO_0002182) using "is measuregroup of" (bao:BAO_0000426) predicate. Links to relevant molecules and membranes are provided in the same ways as a parental bioassay.

Measure groups have assigned methods linked via "has assay method" (bao:BAO_0000212) predicate. All methods are of "assay method component" (bao:BAO_0002753) type and are also assigned to more specific method types defined by MolMeDB vocabulary.

Experimental conditions are described by triples with repr:hasExperimentalCondtion predicate and objects are typed terms of either "molecular charge" (sio:CHEMINF_000120) or "temperature" (efo:EFO_0001702) type (described further in the dataset).

Some interactions are commented on with a detailed description of the interaction. The comment is expressed as a link between the measure group and a string using rdfs:comment predicate.

Measured quantities and their values are expressed as description of endpoints, where each endpoint corresponds to the quantity set by its type and a value. Each measure group is linked to one or multiple endpoints using "has endpoint" (bao:BAO_0000208) predicate.

Table 7. Types of links describing molecule-membrane interaction bioassay measure groups.

Experimental conditions

As mentioned above, experimental conditions are of either "charge" (sio:CHEMINF_000120) or "temperature" (efo:EFO_0001702) type. Charge describes the molecular charge of measured substance and its value is expressed as a triple using "has value" (sio:SIO_000300) predicate. The assigned object is a string literal. Temperature for which the measurement was conducted is expressed as an integer literal linked by "has value" (sio:SIO_000300) predicate. All temperatures are in degrees Celsius (obo:UO_0000027) set by triple with "has unit" (sio:SIO_000221) predicate.

Table 8. Types of links describing molecule-membrane interaction bioassay experimental condition.

Membrane interaction endpoints

The results of bioassay is described by endpoint linked back to its parental measure group by "is endpoint of" (bao:BAO_0000559) predicate. Each endpoint has type asserting measured quantity. Its value is expressed using triples with "has value" (sio:SIO_000300) predicate and a literal object of float type.

Units are terms from ontologies, particular to each endpoint type. Endpoint is linked to its unit by "has unit" (sio:SIO_000221) predicate.

Some endpoints have standard deviations values of which are linked to the endpoint via mmdbvoc:hasStDev predicate.

Table 9. Types of links describing molecule-membrane interaction bioassay endpoints.

Membranes and methods

Membranes (type mmdbvoc:MembrameModel) and assay methods (bao:BAO_0002753) are described similarly. Both are linked by rdf:label predicate to their name, which is string literal. Each membrane and method has multiple other types, which are subclasses of a parent type, to help querying across different categories. They are also linked to their further description using dc:description predicate. Description is a natural (English) language literal of HTML type. Citations for descriptions are provided as links from membrane or method to the citation by dcterms:bibliographicCitation predicate. Citation is also a natural (English) language literal of HTML type. Some membranes are also linked to their molecular components in ChEBI RDF database using "has part" (bao:BAO_0090004) predicate.

Table 10. Types of links describing membranes and measurement or computation methods.

Transporter interactions

Transporter interactions use mmdbtra: namespace and are of "transporter assay" (bao:BAO_0003008) type. Their description is similar to membrane interactions, where each assay is linked to a single measure group (type bao:BAO_0000040) using "has measuregroup" (bao:BAO_0000209) predicate.

Links to screened molecules (type bao:BAO_0000076) and relevant transporters (type bao:BAO_0000283) are provided by triples using "has participant" (bao:BAO_0090012) predicate. Both molecule and transporter protein are further described in the dataset.

The data source is expressed using cito:citesAsDataSource predicate, which links to either a publication (typedcterms:BibliographicResource) or a database (sio:SIO_000089); both are further described in the dataset.

Table 11. Types of links describing molecule-transporter interaction bioassays.

Transporter interaction measure groups

Links from the measure group (type bao:BAO_0000040) provide a detailed description of the measurement. They are linked to parental bioassay (type bao:BAO_0003008) using "is measuregroup of" (bao:BAO_0000426) predicate.

Links to relevant molecule and transporter are provided in the same ways as with transporter assay.

Some interactions are commented on with a detailed description of the interaction. The comment is expressed as a link between the measure group and a string using rdfs:comment predicate.

Measured quantities and their values are expressed as descriptions of endpoints, where each endpoint corresponds to the quantity set by its type and value. Each measure group is linked to one or multiple endpoints using "has endpoint" (bao:BAO_0000208) predicate.

Table 12. Types of links describing molecule-transporter interaction bioassay measure groups.

Transporter interaction endpoints

Assay results are represented by the concentration or categorical endpoint. Both kinds of endpoints are linked back to their parental measure group (type bao:BAO_0000040) using "is endpoint of" (bao:BAO_0000559) predicate. Concentration endpoint are of types "pEC₅₀" (bao:BAO_0000199), "pIC₅₀" (bao:BAO_0002583), "pK_i" (bao:BAO_0190004) and "pK_m"(mmdbvoc:PKm).

Endpoints are connecterd to their values by "has value" (sio:SIO_000300) predicate and are expressed using mol per liter unit (obo:UO_0000062). Some concetration endpoints have standard deviations linked via mmdbvoc:hasStDev predicate.

Table 13. Types of links describing concentration endpoints.

Categorical endpoint states whether a molecule interacts with the target transporter as an inhibitor, non-inhibitor (type mmdbvoc:InhibitionAssay), substrate or non-substrate (type mmdbvoc:SubstrateBindingAssay). They are linked to the result by "has value" (sio:SIO_000300) predicate. The value of the result is either positive (repr:PositiveResult - substance is substrate/inhibitor of a transporter) or negative (repr:NegativeResult  - substance is not substrate/inhibitor of a transporter).

Table 14. Types of link describing categorical endpoints.

Transporters

Targeted transporters are of transporter (bao:BAO_0000283) type. Transporters are linked to their HGNC name as string literal by rdfs:label predicate and is linked to its UniProt ID (edam:data_2291) using "hast attribute" (sio:SIO_000008) predicate. Direct link to equivalent protein record in Uniprot RDF is provided via skos:exactMatch predicate.

Table 15. Types of link describing transporter proteins.

Table 16. Links describing transporter UniProt identifiers

References

References use mmdbref: namespace. Data about molecule-membrane and molecule-transporter interactions in MolMeDB are taken from publications, online databases and in-house calculations (see What is stored in MolMeDB?). Publications (typed dcterms:BibliographicResource) are linked to doi and/or PubMed id (pmid) identifier by "bibo:doi" and/or "bibo:pmid" predicate, respectively. Both doi and pmid are string literals.

Databases (type sio:SIO_000089 "dataset") are linked to their URL using rdfs:seeAlso predicate and assays from in-house calculations have MolMeDB database as their source. Both publications and databases are cited by triples using dcterms:bibliographicCitation predicate. Citation is a string literal.

Both are also linked to relevant interactions (types bao:BAO_0002182 and bao:BAO_0003008) by cito:isCitedAsDataSourceBy predicate and to relevant substances (type bao:BAO_0000076), membranes (type mmdbvoc:MembraneModel) and transporters  (type bao:BAO_0000283) by dcterms:subject predicate.

Table 17. Types of link describing data sources of bioassays in MolMeDB.