Towards a Top-Domain Ontology Towards a Top-Domain Ontology for Linking Biomedical Ontologiesfor Linking Biomedical Ontologies

Holger StenzhornHolger Stenzhorna,ba,b

Elena BeißwangerElena Beißwangercc

Stefan SchulzStefan Schulzaa

aa Department of Medical Informatics, Freiburg University Hospital, Department of Medical Informatics, Freiburg University Hospital, GermanyGermany

bb Institute for Formal Ontology and Medical Information Science Institute for Formal Ontology and Medical Information Science (IFOMIS), Saarbrücken, Germany(IFOMIS), Saarbrücken, Germany

cc Jena University Language and Information Engineering (JULIE) Jena University Language and Information Engineering (JULIE) Lab, GermanyLab, Germany

IntroductionIntroduction

• Tremendous increase in collected data in life Tremendous increase in collected data in life sciences in the last years – especially in sciences in the last years – especially in biomedicine with its genomics researchbiomedicine with its genomics research

• Obvious need to analyze and structure this Obvious need to analyze and structure this gained data to extract new knowledge from itgained data to extract new knowledge from it

• Our focus is on the scientific efforts to use Our focus is on the scientific efforts to use ontologies to achieve this particular taskontologies to achieve this particular task

Problems with Current Ontologies Problems with Current Ontologies

• Most existing biomedical ontologies are Most existing biomedical ontologies are developed mostly independently – even if developed mostly independently – even if they have overlapping content they have overlapping content

• Each ontology only embraces a distinct Each ontology only embraces a distinct scenario with a mere partial view of the fieldscenario with a mere partial view of the field

• An overarching resource has been missing to An overarching resource has been missing to link and interface the independent ontologieslink and interface the independent ontologies

• This could help to use the single ontologies This could help to use the single ontologies more efficiently and create synergetic effectmore efficiently and create synergetic effect

Goals of Our WorkGoals of Our Work

• Creation of the top-domain ontology Creation of the top-domain ontology BioTopBioTop to achieve this interfacing and linking task to achieve this interfacing and linking task

• Concentration upon a smaller selection of Concentration upon a smaller selection of ontologies from the ontologies from the Open Biomedical Open Biomedical OntologiesOntologies ( (OBOOBO)…)…

• e.g. e.g. Gene OntologyGene Ontology ( (GOGO), ), Foundational Foundational Model of AnatomyModel of Anatomy ( (FMAFMA), ), Chemical Chemical Enties of Biololgical InterestEnties of Biololgical Interest ( (ChEBIChEBI))

• Fulfil the principles of Fulfil the principles of OBOOBO in order to join it in order to join it

Basis of Our WorkBasis of Our Work

• BioTopBioTop is based on is based on GENIAGENIA which is a small which is a small ontology (45 classes) that has been…ontology (45 classes) that has been…

• developed in the biological natural developed in the biological natural language processing (BioNLP) community language processing (BioNLP) community

• is intended to be a formal model of cell is intended to be a formal model of cell signalling reactions in humanssignalling reactions in humans

• is limited to highly general classes on is limited to highly general classes on biochemical substances and their location biochemical substances and their location in organismsin organisms

Basis of Our WorkBasis of Our Work

• BioTopBioTop we tried to fix problems found within we tried to fix problems found within GENIAGENIA such as… such as…

• Missing or unclear definitions and Missing or unclear definitions and documentationdocumentation

• Taxonomy sometimes contradicting Taxonomy sometimes contradicting biological or ontological intuitionsbiological or ontological intuitions

• A complete analysis of the found drawback A complete analysis of the found drawback can be found in our AMIA 2006 papercan be found in our AMIA 2006 paper

Ontology TypesOntology Types

• Top Ontology – contains only very small set Top Ontology – contains only very small set of highl-level, general classes (e.g. of highl-level, general classes (e.g. „Continuant“, „„Continuant“, „FunctionFunction“ or „“ or „ObjectObject“)“)

• Top-Domain Ontology – holds essential core Top-Domain Ontology – holds essential core domain classes to interface top and domain domain classes to interface top and domain ontologies (e.g. „ontologies (e.g. „OrganismOrganism“ or „“ or „CellCell“)“)

• Domain Ontology – has a multitude of low-Domain Ontology – has a multitude of low-level, domain-specific classes to describe a level, domain-specific classes to describe a domain of interest (e.g. „Antisense RNA domain of interest (e.g. „Antisense RNA Transcription“) Transcription“)

Ontology TypesOntology Types

Some OBO PrinciplesSome OBO Principles

• BioTopBioTop has a clearly defined topic of being a has a clearly defined topic of being a topic-domain ontology and contains only the topic-domain ontology and contains only the classes to fulfil its taskclasses to fulfil its task

• The development is a collaborative effort The development is a collaborative effort between several institutions involving other between several institutions involving other experts via e.g. face-to-face discusssions or experts via e.g. face-to-face discusssions or mailing listsmailing lists

• It is completely open-source and freely It is completely open-source and freely available to everybody from its websiteavailable to everybody from its website

Integration of Top OntologyIntegration of Top Ontology

• Initial Initial BioTopBioTop version was based on out own version was based on out own set of top-level classes based both on set of top-level classes based both on BFOBFO and and DOLCEDOLCE

• When When BFOBFO became available in OWL-DL we became available in OWL-DL we imported this into imported this into BioTopBioTop

• Integration was straighforward for classes Integration was straighforward for classes taken directly from taken directly from BFOBFO

• For classes taken from For classes taken from DOLCEDOLCE some some remodeling took place via introducing remodeling took place via introducing mediator classesmediator classes

Some OBO PrinciplesSome OBO Principles

• We include precise textual definitions to…We include precise textual definitions to…

• avoid ambiguities many biomedical terms avoid ambiguities many biomedical terms possesspossess

• make the ontology understandable for make the ontology understandable for humanshumans

• It is implemented in a common language It is implemented in a common language (OWL-DL) that allows the use of standard (OWL-DL) that allows the use of standard tools and applicationstools and applications

Example Links to OntologiesExample Links to Ontologies

Biological ProcessBiological Process Biological Process (GO)Biological Process (GO)

Protein FunctionProtein Function Molecular Function (GO)Molecular Function (GO)

Cell ComponentCell Component Cell Component (GO)Cell Component (GO)

CellCell Cell (FMA)Cell (FMA)

AtomAtom Atoms (ChEBI)Atoms (ChEBI)

Subatomic ParticleSubatomic Particle Elementary Particles Elementary Particles (ChEBI)(ChEBI)

Organic CompoundOrganic Compound Organic Molecular Entities Organic Molecular Entities (ChEBI)(ChEBI)

TissueTissue Tissue (FMA)Tissue (FMA)

Example Links to OntologiesExample Links to Ontologies

Usage Scenarios and FutureUsage Scenarios and Future

• BioTopBioTop is used in the EU project is used in the EU project BOOTStrepBOOTStrep for various NLP purposes for various NLP purposes

• classes are used to semantically annotate classes are used to semantically annotate named entities automatically identified in named entities automatically identified in literature abstractsliterature abstracts

• BioTopBioTop as basis for the developement of new as basis for the developement of new ontologies in the framework of ontologies in the framework of OBOOBO and and elsewhereelsewhere

• Continuous amelioration and expansion of Continuous amelioration and expansion of the existing interface classes to expressed the existing interface classes to expressed needs by usersneeds by users

Related WorkRelated Work

• Currently two other top-domain ontology for Currently two other top-domain ontology for biology/biomedicine are worked upon, i.e. biology/biomedicine are worked upon, i.e. Simple Bio Upper OntologySimple Bio Upper Ontology and and GFO-BioGFO-Bio

• To have several top-domain ontologies is To have several top-domain ontologies is counterproductive to their actual goalcounterproductive to their actual goal

• Therefore we want to collaborate with the Therefore we want to collaborate with the other two to achieve a unified ontologyother two to achieve a unified ontology

• work together with the author of the first work together with the author of the first ontology (Alan Rector) is already ongoingontology (Alan Rector) is already ongoing

WebsiteWebsite

• All BioTop material (including the latest All BioTop material (including the latest version of its OWL-DL implementation) is version of its OWL-DL implementation) is available from its websiteavailable from its website

www.ifomis.org/biotopwww.ifomis.org/biotop

The EndThe End