| Metarel |
|
|
|
IntroductionThe Resource Description Framework (RDF) and the Open Biomedical Ontolgy Format (OBOF) share a common basis in their representation models. Both models connect resources or terms trough properties or relation types. Their common basis model has the advantage of simplicity, but lacks a high expressivity, like the Web Ontology Language (OWL). Still, their are many relational features that can be expressed, apart from only a label with a name. Transitivity, reflexivity, subsumption and compositionality of relations need to be used to exploit this model maximally. On the other hand, a carefull distinction is needed between properties that connect class resources and properties that connect instance resources. This distinction allows to make definitions for class level properties by reusing definitions of instance level properties.MetarelMetarel is framework that allows to express the semantics of any properties (rdf:Property in OWL parlance, predicate in RDF parlance and relationship type in the bio-ontology parlance of OBO, examples: 'is part of', 'is located'), also properties that stand in between two classes or between an instance and a class. OWL deals only with the semantics of properties between instances (called owl:ObjectProperty). Apart from the different treatment for properties between classes, Metarel is maximally compatible with OWL. Metarel.rdf is the name of the metagraph for the properties. It allows to add existing properties in its hierarchy and to make appropriate meta-relations between them. Metarel assumes the labeled directed graph model that is common for RDF and OBOF: the nodes in the graph are resources or terms and the directed relation arcs are RDF triples. The nodes have a Unique Resource Identifier (URI) and they can be both instances or classes. A formal classification of properties follows from this graph model. The core of Metarel.rdf is developed as an OBOF formatted ontology. It is exported from there to a representation in RDF, which is also valid OWL Full.
Reasoning on bio-ontologiesBiomedical ontologies are often not formatted in OWL and many of them, especially OBO ontologies, use direct properties between classes (e.g. heart (the class of all hearts) 'is part of' body (the class of all bodies), meaning that all hearts are part of some body). Metarel can be used for semi-automated reasoning with such knowledge statements in large RDF stores. The preparation of the reasoning process requires that the properties used in the RDF store are merged into metarel.rdf. We did this for BioGateway, an RDF store containing many biomedical ontologies. The properties in BioGateway are contained in Biorel, which is a curated extension of OBO's Relationship Ontology (RO). Biorel is integrated in Metarel, forming BioMetarel. The integrated ontology BioMetarel can be readily used for KM tasks in BioGateway. The reasoning steps are executed in a semi-automated way in BioGateway through SPARQL/Update queries. They generated about 158 million new knowledge statements of the form class - property - class. The version of BioGateway with the original (explicit) knowledge statements is gathered in the RDF graph SSB, the inferred knowledge is gathered in SSB_tc. Queries can be launched in the query endpoint of BioGateway. The detailed approach can be followed through the sql files that are used to create BioGateway and the SSB and SSB_tc graphs. The following steps are executed sequentially:
DownloadThe relation metagraph metarel.rdf is a small file that can be downloaded here: The core of metarel in OBOF: The curated properties (relation types) in OBOF: The curated relation types in OWL2 DL: The endproduct for reasoning in RDF:
Publications about Metarel
|
| Last Updated on Thursday, 07 April 2011 16:30 |
New release of ONTO-PERL ONTO-PERL 1.32 features multiple enhancements |
'TurboOrtho - a High Performance Alternative for OrthoMCL' presented at ECCB 2010 as a poster (Ekseth et al.) (The software will be released in April 2011) |
| 'Benchmarking triple stores with biological data' presented at SWAT4LS 2010 as a full paper (Mironov et al.) |