The Ontology of Craniofacial Development and Malformation

American Journal of Medical Genetics Part C (Seminars in Medical Genetics)

A R T I C L E

The Ontology of Craniofacial Developmentand Malformation for TranslationalCraniofacial ResearchJ.F. BRINKLEY,* C. BORROMEO, M. CLARKSON, T.C. COX, M.J. CUNNINGHAM,L.T. DETWILER, C.L. HEIKE, H. HOCHHEISER, J.L.V. MEJINO, R.S. TRAVILLIAN,AND L.G. SHAPIRO

James F. BrInformatics, anbiomedical inf

Charles Borfor the FaceBa

Melissa Clabiology, informthe OCDM.

Timothy C.(Division of Crresearch intere

Michael L.Pediatrics andchildren with c

Landon T. Ddeveloper. He

Carrie L. HeInstitute Crani

Harry S. Hocscientist with e

Jose L.V. MHe is the primMalformation

Ravensara Swith interest in

Linda G. Shwith numerou

Grant spon*Correspon

Seattle, WA. EDOI: 10.100Article first

� 2013 Wil

We introduce the Ontology of Craniofacial Development and Malformation (OCDM) as a mechanism forrepresenting knowledge about craniofacial development and malformation, and for using that knowledge tofacilitate integrating craniofacial data obtained via multiple techniques frommultiple labs and atmultiple levels ofgranularity. The OCDM is a project of the NIDCR‐sponsored FaceBase Consortium, whose goal is to promote andenable research into the genetic and epigenetic causes of specific craniofacial abnormalities through the provisionof publicly accessible, integrated craniofacial data. However, the OCDM should be usable for integrating anyweb‐accessible craniofacial data, not just those data available through FaceBase. The OCDM is based on theFoundational Model of Anatomy (FMA), our comprehensive ontology of canonical human adult anatomy, andincludes modules to represent adult and developmental craniofacial anatomy in both human and mouse,mappings between homologous structures in human and mouse, and associated malformations. We describethese modules, as well as prototype uses of the OCDM for integrating craniofacial data. By using the terms fromthe OCDM to annotate data, and by combining queries over the ontology with those over annotated data, itbecomes possible to create “intelligent” queries that can, for example, find gene expression data obtained from

inkley, M.D., Ph.D. is Professor in Biological Structure at the University of Washington (UW), founder of the field of Structurald director of the Structural Informatics Group (SIG) at the UW. He is an informaticist, with extensive experience in the development oformation systems. He is particularly interested in management, integration and visualization of biomedical data.romeo is a staff programmer in the Department of Biomedical Informatics, University of Pittsburgh. He is a principal software developerse Hub.rkson, M.A., MDes is a Ph.D. student in the UW Biomedical and Health Informatics program. Ms. Clarkson has a background ination design, and scientific illustration. Her dissertation work focuses on the visualization of biomedical knowledge networks such as

Cox, Ph.D. is Professor and Laurel Endowed Chair in Craniofacial Research in the University of Washington's Department of Pediatricsaniofacial Medicine) and the Center for Developmental Biology and Regenerative Medicine at Seattle Children's Research Institute. Hissts lie in using animal models to understand genetic and epigenetic contributions to craniofacial development and dysmorphism.Cunningham, M.D., Ph.D. is the Jean Renny Professor of Craniofacial Medicine in the University of Washington Department ofMedical Director of the Seattle Children's Craniofacial Center. For over 20 years Dr. Cunningham's career has focused on the care ofraniofacial conditions as well as basic and translational research to identify causes and treatments.etwiler, M.S. is a Research Scientist in the UW Department of Biological Structure. Mr. Detwiler is a computer scientist and softwareis the primary technical maintainer and developer of the FMA, OCDM, and Query Integrator tool.ike, M.D., M.S. is Assistant Professor in the UW Department Pediatrics. Dr. Heike is a member of the Seattle Children's Researchofacial Center and the Center for Clinical and Translational Sciences. She is an expert in the clinical aspects of craniofacial disorders.hheiser, Ph.D. is Assistant Professor in the University of Pittsburgh Department of Biomedical Informatics. Dr. Hochheiser is a computerxpertise in visualization and human computer interaction. He is the technical director of the FaceBase Hub.ejino, M.D. is Senior Research Scientist in the UW Department of Biological Structure. Dr. Mejino is an anatomy and ontology expert.ary author and curator for the Foundational Model of Anatomy Ontology (FMA) and the Ontology of Craniofacial Development and(OCDM), as well as a contributor to many other ontologies and ontology best practices.. Travillian, Ph.D. is a postdoctoral Fellow in the UW Department Computer Science and Engineering. Dr. Travillian is an informaticist,comparative anatomy. She is the primary developer of the mouse components of the OCDM.

apiro, Ph.D. is Professor in Computer Science and Engineering (CSE) at the UW. She is an expert in imaging, graphics, and visualization,s publications in the area of computer vision, pattern recognition, ontology, data integration and content‐based retrieval.sor: NIH FaceBase Consortium; Grant number: U01DE020050.dence to: J.F. Brinkley, M.D., Structural Informatics Group, Department Biological Structure, Box 357420, University of Washington,‐mail: [email protected]/ajmg.c.31377published online in Wiley Online Library (wileyonlinelibrary.com).

ey Periodicals, Inc.

2 AMERICAN JOURNAL OF MEDICAL GENETICS PART C (SEMINARS IN MEDICAL GENETICS) ARTICLE

mouse structures that are precursors to homologous human structures involved inmalformations such as cleft lip.We suggest that the OCDM can be useful not only for integrating craniofacial data, but also for expressing newknowledge gained from analyzing the integrated data. © 2013 Wiley Periodicals, Inc.

KEYWORDS: ontologies; data integration; craniofacial development and malformation; knowledge representation

How to cite this article: Brinkley JF, Borromeo C, Clarkson M, Cox TC, Cunningham MJ, Detwiler LT, HeikeCL, Hochheiser H, Mejino JLV, Travillian RS, Shapiro LG. 2013. The ontology of craniofacial developmentand malformation for translational craniofacial research. Am J Med Genet Part C Semin Med Genet

9999:1–14.

FaceBase is a consortium ofbiology and technology

projects, the purpose of whichis to systematically acquireand integrate multiple formsof data from human and

model organisms in order tofacilitate research aimed at asystems level understanding of

the causes and possibletreatments for craniofacial

abnormalities.

INTRODUCTION

Craniofacial abnormalities compriseover half of all congenital malforma-tions. Even for the most common groupof craniofacial malformations, orofacialclefts, the etiology is unknown for mostaffected individuals, the phenotypicvariability is great, and the treatmentoutcomes vary considerably. These gapsin knowledge impact our ability tocounsel patients regarding recurrencerisks, and to optimize managementpractices and long‐term treatment out-comes. Even for well‐established syn-dromes associated with clefting withknown genetic mutations, the pheno-typic variability can significantly impactclinical management.

For example, the well‐recognizedphenotype associated with Apert syn-drome [Robin et al., 1993; OnlineMendelian Inheritance in Man, 2010]is characterized by craniosynostosis,midface hypoplasia, syndactyly, and cleftpalate. It is almost exclusively caused byone of two point mutations in thefibroblast growth receptor 2 (FGFR2)gene [Wilkie et al., 1995]; yet pheno-typic variation can significantly affectclinical outcome among patients withthe same condition. For example, someaffected individuals have relatively mildmidface hypoplasia with minimal impacton respiratory status, while other indi-viduals have such severe midface hypo-plasia they require significant surgicalintervention. Identification of the ge-netic and epigenetic factors that deter-mine sub‐phenotypes in this or othercraniofacial conditions would lead to abetter understanding of pathogenesis,allow clinicians to tailor treatment tooptimize outcomes, enable further re-search into possible prevention andtherapeutic strategies, and facilitatemore effective counseling for families.

To understand the genetic andepigenetic contributions to the clinicalvariability of craniofacial conditionsrequires precise and detailed phenotypicand genotypic data from multiplesources. This may include temporallyrestricted gene expression data frommicroarray or RNAseq experiments,and 3D shape information from stereo-photogrammetry or Computed Tomog-raphy, acquired by multiple labs aroundtheworld. However, until recently, therewas no systematic method for integrat-ing all these data in order to morecompletely and precisely characterizecraniofacial phenotypes.

Recognizing this problem, theNational Institute of Dental and Cra-niofacial Research (NIDCR) sponsoredthe creation of FaceBase in 2009. Face-Base is a consortium of biology andtechnology projects, the purpose ofwhich is to systematically acquire andintegrate multiple forms of data fromhuman and model organisms in order tofacilitate research aimed at a systems level

understanding of the causes and possibletreatments for craniofacial abnormali-ties. The FaceBase Hub coordinatesconsortium activities and provides datato the research community through apublic web repository [Hochheiser et al.,2011].

THE BENEFITS OF ANONTOLOGY FOR DATAINTEGRATION

In order to maximize the utility ofexisting and future data in repositoriessuch as FaceBase, it is essential to link thedata through standardized terminology[Olson et al., 2008]. Once data areannotated (or associated) with commonterminology, they may be integrated bysearching for common terms, in amanner similar to a Google search.However, such keyword‐based searchesstill require considerable user knowl-edge, since only the investigators knowthe associations between terms. Toachieve integration at the level ofmeaning (semantics) rather than simpletermmatching requires that the terms beorganized in an ontology, which amongother things defines specific relationsamong terms [Sowa, 1999].

The importance of ontologies toaccomplish this type of semantically‐basedintegration is highlighted by an activeresearch community in the developmentof both ontological content and compu-tational techniques for using ontologies,often in conjunction with genetic andgenomic data, to derive novel insights.The National Center for BiomedicalOntology provides access to severalhundred such ontologies through theBioportal archive [Whetzel et al., 2011;Musen et al., 2012]. Two of the manyexamples of the benefits of ontologiesinclude using the Gene Ontology

The importance of ontologiesto accomplish this type of

semantically‐based integrationis highlighted by an activeresearch community in the

development of bothontological content and

computational techniques forusing ontologies, often in

conjunction with genetic andgenomic data, to derive novel

insights.

ARTICLE AMERICAN JOURNAL OF MEDICAL GENETICS PART C (SEMINARS IN MEDICAL GENETICS) 3

[Ashburner et al., 2000] to identifyclasses of genes that may be over‐represented in expression data sets[Shah et al., 2012] and using cross‐species phenotype descriptions, poten-tially in combination with genomic data,to infer relationships between humandiseases and animal models [Washingtonet al., 2009; Bauer et al., 2012; Chenet al., 2012; Doelken et al., 2012].

WHY A DEDICATEDONTOLOGY FOR THECRANIOFACIALCOMMUNITY?

Many existing ontologies include termsthat are relevant for craniofacial research,such as the Foundational Model ofAnatomy Ontology (FMA) [Rosse andMejino, 2003, 2007] for canonical adulthuman anatomy; and the Human De-velopmental Anatomy [Bard, 2012] forembryological development; SNOMED‐CT [Spackman and Campbell, 1998],the Human Phenotype Ontology[Robinson and Mundlos, 2010] DiseaseOntology [Du et al., 2009] and Elementsof Morphology [Allanson et al., 2009]for phenotypic abnormalities; and thePhenotype Trait Ontology [Mungallet al., 2010] for description of pheno-typic qualities.

In addition, translating betweenhuman anatomy, development, and

malformation and both analogous andhomologous counterparts in other spe-cies is an important goal for the researchcommunity. Ontologies for adult mouseanatomy [Hayamizu et al., 2005]; devel-opmental mouse anatomy [Baldocket al., 1999; Armit et al., 2012], andmammalian phenotype [Smith andEppig, 2012] parallel the human ontol-ogies. A number of efforts have devel-oped cross‐species ontological links forboth anatomic [Travillian et al., 2011b;Dahdul et al., 2012; Hayamizu et al.,2012; Mungall et al., 2012; Niknejadet al., 2012] and phenotypic [Kohleret al., 2013] descriptors.

Despite the breadth of these efforts,these ontologies and mappings currentlydo not provide sufficient anatomical,developmental, and phenotypic detail tomeet all of the needs of the craniofacialresearch community for describing hu-man development and anatomy, and forrelating these descriptions to relevantcounterparts in model organisms such asmice. For example, whereas craniofacialresearch requires a comprehensive re-presentation of head anatomical entitiesand their relations, an ontology such asthe mouse anatomy ontology [Hay-amizu et al., 2005] represents only twotypes of relation (part_of and is_a), withlimited depth for craniofacial structures.Similarly, whereas the mapping ontolo-gy UBERON [Mungall et al., 2012] wasdeliberately designed to be homology‐neutral, craniofacial researchers oftenneed to link orthologous genes (i.e.,inherited from a common ancestor,and distinguished from one anotherthrough speciation events) acrossspecies. It is for these reasons that theNIDCR sponsored the creation of theOCDM.

The purpose of this article is tointroduce theOCDM to the craniofacialresearch community, to describe itscurrent status and future plans, and toprovide examples of how the OCDMwill be of use not only to FaceBase butalso to geneticists, dysmorphologists,and craniofacial investigators. We beginby describing our overall approach, andthen provide specific details on ourdriving use cases, the organizationalframework and content entered to‐

date, and prototype examples of thepotential use of the OCDM.

OVERALL APPROACH

When complete, the OCDM willinclude normal and abnormal cranio-facial structure and function, develop-mental anatomy and processes,syndromes, and other conditions rele-vant to craniofacial research and clinicalpractice. It will include links to corre-sponding model organism knowledgesources, as well as relevant genomicknowledge sources such as the geneontology [Ashburner et al., 2000].

To provide a framework for thislarge and diverse amount of informationwe organize the OCDM primarilyaround human adult and developmentalanatomy, as provided by our Founda-tional Model of Anatomy (FMA) [Rosseand Mejino, 2003]. The FMA is asymbolic representation of the pheno-typic structure of the body ranging frommacroscopic anatomy to molecules,which is by design limited to canonicalstructure. The FMA includes a level ofdetail that is not found in anatomytextbooks like Gray’s Anatomy[Gray, 1918]. For example, as shown inFigure 1, in order to establish ontologicalrelationships between the adult lip, acleft lip, embryologic development ofthe lip, and gene expression in the lipregion, the FMA includes definitionsfor many subcomponents and bounds ofthe lip that are unnamed in standardtexts.

The FMA currently comprises over93,000 classes (e.g., “Head,” “Mouth”)represented by about 170,000 termsincluding preferred names, synonyms,eponyms and non‐English equivalents,as well as over 2.3 million relationsbetween classes (is_a, part_of, etc.). TheFMA is currently the largest and mostcomprehensive human anatomy ontol-ogy, and is becoming widely accepted asthe canonical anatomy reference ontol-ogy, with most of the major ontologiesand controlled terminologies, suchas SNOMED‐CT [Spackman andCampbell, 1998], RadLex [Kunduet al., 2009], and Terminologia Ana-tomica [Whitmore, 1999] aligning to or

Upper lip

Left sideof upper lip

Right sideof upper lip

Philtrum

Right upper lip proper Vermilion of right sideof upper lip

Left upper lip proper Vermilion of left sideof upper lip

Philtrum proper Vermilion of philtrum

Figure 1. Schematic diagram of the regional parts of the upper lip.

The FMA is currentlythe largest and mostcomprehensive human

anatomy ontology, and isbecoming widely accepted as

the canonical anatomyreference ontology, with mostof the major ontologies andcontrolled terminologies, suchas SNOMED‐CT aligning toor incorporating the FMAas the anatomy axis oftheir own ontologies.


incorporating the FMA as the anatomyaxis of their own ontologies.

Given this anatomical frameworkwe are building the OCDM by firstenhancing the FMA to provide detail onboth adult and developmental craniofa-cial anatomy as needed by a series of usecases (described in the next section) thatwe have elicited from craniofacial re-searchers. We then extract a view (or“slim” in GO terms) [Gene OntologyConsortium, 2013] of just the craniofa-

cial components of the FMA neededto create the canonical craniofacialhuman sub‐ontology (CHO), and uti-lize this view as a template for develop-ing other OCDM components (or sub‐ontologies) dealing with human malfor-mations, normal and abnormal mousedevelopment and malformations, andmappings between homologous ana-tomical structures in human and mouse.Our current focus is on human andmouse, and a similar approach ofmapping to canonical human anatomyas the fundamental organizing principlecould be used with other relevant modelorganisms.

Content creation for the OCDM isdone in the Protégé ontology authoringtool [Gennari et al., 2003], and is madeavailable as a semantic web service[McIlraith et al., 2001], which allowsthe ontology to be queried by end‐userapplications. Whereas the OCDMcontent and the web services are“under‐the‐hood,” of interest mainlyto ontology authors, it is the end‐userapplications that will provide the func-tionality and graphical interfaces thatallow craniofacial researchers to utilizethe OCDM in their own work.

In the following sections we providemore detail on the use cases, the currentOCDM content, example queries thatcan currently be answered by the

OCDM, and a prototype applicationthat accesses the OCDM via the webservice.

USE CASES

The development of the OCDM isinformed by a series of use cases, wherefor our purposes a “use case” is a task orset of tasks that a craniofacial researcherwould carry out in his or her work,and that would use the OCDM. Basedboth on our experience in buildingthe FaceBase Hub and on input fromcraniofacial researchers in the consor-tium, these use cases describe tasks andscenarios that might use OCDM‐baseddescriptions to support data navigation,exploration, and interpretation.

Each use case is characterized by adescription, a list of potential users (e.g.,clinician, researcher), a list of tasks thatneed to be accomplished to implementthe use case, a list of information needsthat should be met by the ontology, thedata or both; and a list of user interfaceissues that need to be addressed in orderto make the implemented use caseaccessible to craniofacial researchers.Current use cases are organized in fivecategories—data annotation, searching/browsing, visualization, gene expressiondisplay, and analytics (Table I).

TABLE I. Categories of Use Cases

Data annotation Associating OCDM terms with images or other data that are uploaded to the FaceBase Hub. The annotations canbe associated with sets of related data files, with individual files or even subsets, such as selected regions inimages

Searching/Browsing Using terms and relations from the OCDM to search for data, and to navigate between data points based onrelations

Visualization Graphical and navigable displays of ontological concepts and relations. Possibilities includes graph‐based browsingand reference atlas displays that depict anatomic or phenotypic characteristics directly on 2D or 3D renderingsof human or animal images

Gene expressiondisplay

Organize gene expression results based on anatomic or phenotypic annotations from the OCDM, along withother metadata such as developmental stage and species. Results might be summarized automatically in amanner similar to the Gene Wiki [Huss et al., 2010] or contrasted across closely related samples (e.g., adjacentanatomic regions at a common developmental time point)

Analytics Support enrichment analyses or other data mining algorithms that use ontological structure to infer relationships.Examples include finding associations between genes and specific precursor structures associated with amalformation, and cross‐species phenotype analyses [Washington et al., 2009; Bauer et al., 2012; Chenet al., 2012; Doelken et al., 2012]


As one specific example Table IIdescribes the use case, Finding structuresassociated with syndromes, which is classi-fied under “Searching/Browsing.” In

Finding structures associated with syndromesA clinician or researcher is interested in le

syndrome). Starting from the name of tFor each associated structure:

Find the homologous structure in mousFind developmental progenitors—earlyFind associated gene expression data thaOrganize results by structure and develo

The inverse—finding syndromes associatedRoles

ClinicianTranslational researcher

TasksMapping syndrome descriptions to affecExamining corresponding anatomic stru

Information needsAssociation of malformation syndromesLinks between human and animal anatoAnatomic precursors and successorsLinks between structures and gene exprAnnotation of structures with developm

Interface implicationsGraphical anatomic region selectionDevelopmental stage timelineSet‐based relationsDynamic data tables

this scenario, a user is interested infinding relevant clinical or basic researchdata that might be relevant to Apertsyndrome. Rather than having to

TABLE II. Sample Use Case

arning more about anatomical structures in Aperhe syndrome of interest, the clinician identifies a

e or other organismsstructures that led to one or more structures invot’s been uploaded to Facebasepmental stagewith structures—is also possibly relevant

ted structuresctures in developing organisms

with associated anatomic structuresmic structures

ession dataental stages

tediously search through the literatureor online databases the user in thisscenario would simply type “ApertSyndrome” into a user interface. The

t syndrome (or some other craniofacialssociated anatomic structures.

vled in the syndrome of interest


system would then consult the OCDMto find the reported malformationsassociated with Apert. The user mightthen select those malformations relevantto a specific case of interest. The systemwould then find the associated anatomicstructures, the developmental precursorsto those structures, and the homologousstructures in the mouse or other modelorganisms. For each of these structures itwould then consult online databases likeFaceBase, and return a list of publicallyavailable data (such as gene expressiondata) that have been acquired about eachof these structures. At the same time thesystemmight also search Pubmed to findarticles indexed by these structures,pruned by other factors such as speciesand developmental stage.

This and other OCDM use casescan be viewed on the FaceBase Hubat https://www.facebase.org/ocdm/.Registered FaceBase users are invitedto use the Hub’s comment facility todiscuss use cases and suggest additionsand revisions.

Figure 2. OCDM co

OCDM CONTENT

Figure 2 shows that the OCDM isorganized as a series of components(also called sub‐ontologies) that areincluded in the overall ontology. Eachis developed somewhat independently,although coordination is required. Thecurrent sub‐ontologies are:

�

m

Craniofacial Human Ontology (CHO):canonical human adult craniofacialanatomy, derived from the FMA.

�
Craniofacial Human Development Ontolo-gy (CHDO): canonical human develop-mental craniofacial anatomy, derivedfrom the FMA as a subset of the CHO.
�
Craniofacial Mouse Ontology (CMO):canonical mouse adult craniofacial anat-omy, based on the CHO, and augment-ed with mouse‐specific terms.
�
Craniofacial Mouse Development Ontology(CMDO): canonical mouse develop-mental craniofacial anatomy, basedon the CHDO, and augmented withmouse‐specific terms.
ponent ontologies.

�
Craniofacial Human‐Mouse MappingsOntology (CHMMO): structure‐basedmappings for homologous anatomicalentities.
�
Craniofacial Human Malformation Ontolo-gy (CHMO): phenotypic human cra-niofacial malformations.
�
Craniofacial Mouse Malformation Ontology(CMMO): phenotypic mouse craniofa-cial malformations.
�
Craniofacial Human‐Mouse MalformationMappings Ontology (CHM3O): map-pings between human and mousemalformations.
�
Attribute Entity Ontology (AEO): quan-titative and qualitative physical descrip-tors of the attributes and propertiesexpressed in other components of theOCDM.
The initial focus for FaceBase wason the midface and cleft lip/palate (CL/P), so most of the current content of theOCDM to‐date is related to these areas.Expansion to other malformations orsyndromes should be faster in the futurenow that the organizational frameworkis in place. The CHO is nearingcompletion for CL/P, and the CMO,CHMMO and AEO are in activedevelopment. We thus describe thesein more detail in the next sections.The mouse malformations ontology(CMMO) and mappings from humanto mouse malformations (CHM3O) arecurrently only in rudimentary form sowe defer their description to futurereports. Note that the modular organi-zation of the OCDM will permitadditional components to be developedin the future (such as gene to phenotypemappings, other model organisms, orphysiological manifestations keyed tothe associated anatomical entities).

CRANIOFACIAL HUMANONTOLOGY (CHO, CHDO)

The CHO and CHDO are thesubset of the FMA that deals withthe canonical representation of bothadult (CHO) and developmental(CHDO) craniofacial structures. In thefollowing sections, classes (as for exam-ple, anatomical structures) begin with a

Figure 3. Regional (R) andconstitutional (C) parts of the upperlip as shown in our OCDM ontologyviewer. Clicking a horizontal arrowopens up more detail.


capital letter and relations betweenclasses are in italic.

Canonical Adult Anatomy (CHO)

The craniofacial content of the FMAwasextended to include all the spatio‐structural relationships necessary tocomprehensively represent the structuralcomponents involved in the genesis oforofacial clefts. CL/P is the mostcommon type of orofacial cleft, so thecontent enhancement largely focusedon all anatomical entities and structuralrelationships pertaining to the lips,nose, oral vestibule, gingiva, alveolarridge, palate, and orbital structures.

CL/P is the most commontype of orofacial cleft, so thecontent enhancement largelyfocused on all anatomicalentities and structural

relationships pertaining to thelips, nose, oral vestibule,

gingiva, alveolar ridge, palate,and orbital structures.

Figure 4. Class inheritance (is_a)hierarchy of developmental structures.

We augmented the ontological descrip-tion of canonical adult anatomy togranularity levels that closely correspondto developmental information thatspecifies the structures affected duringdysmorphogenesis. For example, a lat-eral cleft upper lip is characterized by agap between the philtrum and a part ofthe upper lip that is unnamed in text-books. We identified the unnamed partsby extending the partition of the upperlip into three major regions: Philtrum inthe middle, flanked by Right side ofupper lip and Left side of upper lip(Fig. 1).

Figure 3, which is generated by ourOCDM ontology viewer, illustrateshow this partition is captured in theontology. In this case, both regional parts(partition by topography, e.g., lateralregion or zone of organ) and constitu-tional parts (partition by composition,

e.g., tissues) are represented. Eachregional part derives from differentembryological structures. A cleft resultsfrom either lack of, or partial fusionbetween the precursor of the philtrum(intermaxillary process) and that ofeither the right or the left side of theupper lip (right or left maxillary promi-nence). Another partition of clinicalimportance is subdividing the upperlip into Vermilion of upper lip andUpper lip proper to distinguish betweentwo constitutionally distinct parts,which can serve as landmarks forincomplete clefts. Hence, we haveformally associated the developmentalstructures with their adult versionsby enhancing the granular partition ofthe upper lip at different stages oftransformation.

Developmental Anatomy (CHDO)

The Craniofacial Human DevelopmentOntology is a subset of the CHO. It is animplementation of the AnatomicalTransformation Abstraction component(ATA) of the FMA [Rosse and Mejino,2007], which was created to provide thehuman development framework for theFMA. The ATA accounts for normalhuman development from the zygote tothe organism’s post‐gestational stage.The high‐level class structure ofthis human developmental ontologyparallels the adult counterpart, where

developmental anatomical entities arerepresented and defined in terms of theirphysical properties, which explicitlyspecify their structural types. Embryo-logical structures are classified accordingto the FMA definitions of the differentclasses of anatomical structures (Fig. 4)such as Organ (e.g. Neural tube is_aEmbryonic organ, which in turn is_aDevelopmental organ), or Organ part(e.g., Rhombomere is_a Region ofneural tube, which in turn is_a Embry-onic organ part).

The ontology framework we havebuilt leverages existing developmentalontologies, such as the EdinburghHuman Developmental Anatomy(EHDA) [Bard, 2012]. However, unlikethe EHDA, we extended the ontologicalrepresentation to other relationshipsbetween the developing structures atdifferent stages of development. Forexample, we instantiated spatial‐struc-tural relationships such as parthood(e.g., Palatopterygoquadrate bar part_ofMaxillary prominence), adjacency(e.g., Cytotrophoblast surrounds Extra‐embryonic mesoblast), connectivity

Figure 5. Schematic diagram ofthe development of the hard palateand upper lip as represented in theOCDM. Grey boxes are classes,labeled lines are relations.

The CraniofacialHuman‐Mouse MappingsOntology (CHMMO)contains structure‐basedmappings for homologous

anatomical entities.CHMMO mappings areexplicit hypotheses of

anatomical homology betweentwo species‐specific structures.


(e.g., Bilaminar embryonic disc continu-ous_with Amnion). In addition, weintroduced temporal properties (Carne-gie stage, gestational age, days post‐ovulation) and developmental properties(transforms, derives, fuses, merges), as forexample, Frontonasal prominence derivesRight nasal placode and Left nasalplacode. Figure 5 shows this relation aswell as many others that allow us torepresent in the OCDM the develop-mental lineage of structures involvedin CL/P. This level of detail, which isnot present in other developmentalontologies, will allow us to identifyproperties (e.g., gene expression) thatare involved in each stage of develop-ment, as well as determine what struc-tures at what stage are affected whenmalformation occurs. The CHO andCHDO together include 11,451 classes,23,389 terms and 196 relations (proper-ties), many of which were alreadypresent in the FMA.

Craniofacial Mouse Ontology(CMO, CMDO)

The Craniofacial Mouse Ontology(CMO) and Craniofacial MouseDevelopment Ontology (CMDO) arerepresentations of canonical mouse cra-niofacial anatomy, both adult and devel-opmental. We used the CraniofacialHuman Ontology (CHO) as a template,pruning and editing as needed torepresent anatomically homologousmouse structures, and then integratedMA and EMAP structures. We usedinference of the existence of missingstructures from the vertebrate Bauplanand FMA, and verified these structures,as well as finding omitted ones, fromdomain experts and the translationalresearch literature. As a result thetaxonomy, property, and annotationproperty structures in the CMO arealmost the same as in the CHO.Additionally, the derivation of mousestructures from existing FMA/CHOstructures necessitated some additionalannotation property slots to provide asound representation. To date there are11,063 classes, 23,383 terms and 198relations in the CMO/CMDO. Ofthese, 482 classes were acquired fromMA or EMAP.

Craniofacial Human‐MouseMappings Ontology (CHMMO)

The Craniofacial Human‐Mouse Map-pings Ontology (CHMMO) contains

structure‐based mappings for homolo-gous anatomical entities. CHMMOmappings are explicit hypotheses ofanatomical homology between twospecies‐specific structures. This was anexplicit design decision, made for thepurpose of enabling linkages to bedrawn among anatomic and phenotypicstructures derived from orthologousgenes.

The CHMMO was built from theground up, although it operates on thesame mapping principles as the Com-parative Anatomy Information System(CAIS) [Travillian et al., 2011b] and theVertebrate Bridging Ontology [Travil-lian et al., 2011a]. Initially, structureswith similar or the same names betweenspecies are provisionally assumed to behomologous. Homologies betweenstructures with dissimilar names acrossspecies are added as needed.

These provisional mappings arevalidated by domain experts. To‐datewe have used a single expert, Dr. Cox(one of the co‐authors); in the future wewill establish a mechanism for others tocomment on the mappings as well asother aspects of the OCDM. To validatethe provisional mappings the domainexpert and content developers consultthe appropriate research literature, as, forexample, Incus (Homo sapiens)$ Incus(Mus musculus), Malleus (Homo sapi-ens)$Malleus (Mus musculus), Stapes(Homo sapiens)$ Stapes (Mus muscu-lus) [Kanzaki et al., 2011]. In generalvalidation is based on explicit molecularevidence when available [Brugmannet al., 2006; Laue et al., 2011]; inthe absence of direct molecular evi-dence, mappings are inferred on thebasis of the vertebrate Bauplan [Depewand Compagnucci, 2008; Fish et al.,2011].

The result of these validations is aconfidence level assigned to the map-ping, together with the provenance ofthe mapping, the expert’s confidencein the mapping, and the number ofreturned results in the literature.

The resulting mappings and confi-dence levels are represented as instancesof classes in the CHMMO, where eachrepresents a type of mapping. Forexample, Alveolar bone of maxilla


maps 1‐1 with Alveolar bone of maxilla(Mus musculus) [355 such mappings].Vibrissa (Mus musculus) is a nullmapping, because there is no suchcorresponding structure in humans [38such mappings], and Snout (Mus mus-culus) maps to a corresponding region ofthe face that is yet to be named(nasomaxillary region) in humans [2such mappings]. An advantage of thesespecific types of mappings is that eventhough there are null mappings such asVibrissa, the OCDM will be able toretrieve the structures from which it isderived, which will have a 1‐1 mapping,thus highlighting points of developmen-tal divergence.

Figure 6. Classes and relations involved in the genesis of cleft upper lip, asrepresented in the OCDM.

Attribute Entity Ontology (AEO)

The AEO provides quantitative andqualitative physical descriptors of theattributes expressed by classes in theOCDM. These are the physical objectproperties of anatomical structures andpathological entities, such as size, shape,distance, and mass. We developed theAEO based on the Phenotypic QualityOntology (PATO), which is designed torepresent “phenotypic qualities” of“quality‐bearing entities” [Mungallet al., 2010]. An example of such anentity is the “vermilion of upper lip”which could be the bearer of the quality“thin.” The combination therefore de-scribes the “thin vermilion” phenotype.This approach allows us to extend thephenotypic descriptions to a level ofdetail that is necessary to sort out sub‐phenotypes and to accommodate quan-titative measurements, without requir-ing an a priori distinction betweennormal and abnormal. Thus, the AEOprovides an ideal framework withinwhich we incorporated the standardsused to describe human morphologythat were created and published by theElements of Morphology WorkingGroup (http://elementsofmorphology.nih.gov/). We are also populating thisontology with other attributes for theanatomical entities related to CL/P.There are currently 106 classes, 119terms, and 7 relations in AEO, of which43 high level classes were derived fromPATO.

Craniofacial Human MalformationOntology (CHMO)

Many cleft classification systems exist[Marazita and Mooney, 2004]. In devel-oping the CHMO for CL/P, we firstidentified and compared over a dozenpublished classification systems for CL/Pphenotypes. Based on this comparison, afull description of which is in prepara-tion, we concluded that there is as yet nouniversal classification system that cap-tures all variant presentations of CL/P.Given the existence of so many ap-proaches, each of which has merit in itsown right, our goal is not to invent yetanother classification scheme but ratherto create a unifying framework that willaccommodate and reconcile existingclassification schemes. To create such aframework requires that we explicitlyrepresent the assumptions underlyingeach classification scheme in terms of thebasic anatomy, pathology and develop-mental processes.

To create this framework we firstconsulted several existing ontologiesor controlled vocabularies relating tomalformations, namely: SNOMED‐CT [Spackman and Campbell, 1998]Human Disease Ontology (HDO) [Duet al., 2009], Human Phenotype Ontol-

ogy (HPO) [Robinson and Mundlos,2010], ICD‐10CM [NCHS, 2013], andElements ofMorphology [Allanson et al.,2009]. In general, most sources do notinclude representation of the underlyingbasic science in the phenotypic descrip-tions. For example, HPO classifies “cleftupper lip” as an abnormality of the upperlip but does not clearly specify whetherthe term pertains to the pathologicalstructure itself or its phenotypic abnor-mality. It is important to distinguishbetween the two because the kinds ofbasic information that are associated witheach are significantly different from oneanother. That is, pathological structuresmay be described by morphometricmeasurements (e.g., upper lip height,philtrum width) whereas phenotypicabnormalities may be described byprocessual properties (e.g., transforms,derives, fusion, or partial fusion).

The framework we have createdtherefore consists of building separatesub‐ontologies for pathological struc-tures and phenotypic abnormalities,using canonical anatomy as the basisfor classifying both types of entities, andontology best practices to ensure consis-tency with high level classificationsystems [Grenon et al., 2004; Rosseet al., 2005].


Figure 6 illustrates this dualrepresentation for cleft upper lip. Be-cause we regard cleft upper lip as apathological structure (as opposed to aphenotypic abnormality) we say thatCleft upper lip is a Pathologicalstructure. We say that Cleft upperlip has_condition Clefting of upperlip, which is a Phenotypicabnormality.

We then associate these entities withtheir corresponding canonical structuresin the CHO. Thus, Cleft upper lip is avariation_of Upper lip, and Clefting ofupper lip is located_in the Upperlip, wherein a gap exists betweencertain regions of the upper lip. Wealso associate morphometric measure-ments and descriptors with canonicalstructures. These measurements, whichcome from the AEO, include suchproperties as “cleft” and length (“phil-trum length”), that are the basis forclassifying an Upper lip as a Pathologicalstructure. Similarly, we plan to associatephenotypic abnormalities such as Cleft-ing of upper lip, with pathologicalprocesses such as Partial‐fusion orNon‐fusion of precursor developmentalstructures.

By associating these types of entitieswith each other and with canonicaldevelopmental structures via specificrelations whose basic biological meaningis well‐defined, we can represent quali-tative knowledge of developmentalbiology in enough detail that we cantrace the morphogenesis of clinicalmalformations, and hence search theappropriate databases for experimentalevidence of genetic or epigenetic factorsthat could be associated with theirpathogenesis. Although these detailedrelations may seem unnecessary to aclinician or researcher, they are neededin order to make the knowledge“computable” so that it can notonly be used for reconciling differentclassification schemes, but also for“intelligent” searches that can traversethese links to find data relevant tomalformations. We illustrate such asearch in the next section. Currentlythere are 280 classes and 21 relations inthe CHMO, with 16 classes derivedfrom HPO.

Prototype Uses

To be of greatest utility the ontologyneeds to be “under the hood,” embed-ded in applications that allow end usersto perform the kind of tasks required bythe use cases described earlier. Theontology could be distributed withvarious applications if it is small enough,or could be made available as a webservice that can accept queries from bothhuman users and software applications.Advantages of this approach are that theontology remains up‐to‐date, and theanswers to queries may more easily becombined with other web‐accessibleontology and data sources.

We are taking the latter approachby periodically making the OCDMavailable as a queryable semantic web[Berners‐Lee et al., 2001] service.Queries over this service are createdand saved in our Query Integratorapplication (QI) [Brinkley andDetwiler, 2012]. TheQI allows differentqueries to be integrated, as for example,a query over the ontology with a queryover data, as we describe in the nextsection. In addition, saved queries can beaccessed by end‐user applications thathide the details of the underlyingontology and query engine.

In the next two sections we describeexamples of saved queries over theOCDM, as well as an example applica-tion that accesses saved queries, butpresents the results in a graphical formthat is more attuned to end users.

Queries

There are currently about 30 savedqueries over the OCDM in the QueryIntegrator database. Links to executableversions of several of these queries can befound on the OCDM page available atthe FaceBase Hub https://www.face-base.org/content/ocdm. For example, aquery on the “human nose” finds theparts of the human nose, and then foreach of these parts, finds the mapping (ifany) to the corresponding homologousstructures in the mouse. Similarly, aquery on the “human right nasal bone”retrieves the facial landmarks associatedwith that bone. These and similar

landmarks will be useful for retrievingspecific measured distances from mor-phometric data, as for example, thenormative data for facial measures thatare currently available through the Face-Base website.

The above queries, however, areonly over the ontology. A third queryintegrates two queries: one over theontology and one over a data source, inorder to perform an “intelligent” query.In this case the data source is theFaceBase Hub, which currently housesover 200 datasets contributed by Face-Base consortium members.

When datasets are uploaded to theHub they are annotated with terms froma set of controlled vocabularies that areaccessible via pull‐down menus. Theseterms can then be used to search forspecific datasets by clicking checkboxesin the search interface shown in Figure 7.One of the controlled vocabularies isAnatomy, which has been clicked toopen up the list of Anatomy terms fromthe OCDM. The user has then checkedMedial nasal process, which causes thesearch engine to retrieve all those data-sets that have been annotated withMedial nasal process, in this case geneexpression data from the developingmouse. A user looking for data relevantto cleft lip might know that the left andright medial nasal processes are develop-mental precursors to the philtrum(Fig. 5), which is in turn part of theupper lip and not fused in some kinds ofcleft lip. But if the user did not have thisknowledge, they would not know toclick on medial nasal process in the Hubsearch. However, this knowledge ispresent in the ontology, so we haveused the Query Integrator to integrate aquery over the ontology with one overthe Hub annotated data, allowing theuser to start just with cleft lip and returnall those datasets annotated with adevelopmental precursor of cleft lip.

The ontology query first examinesrelations like those shown in Figure 6 todetermine that Cleft lip has subclassCleft upper lip, which is a variation_ofUpper lip. The query then examinesrelations like those shown in Figure 5 tofind the developmental lineage of Upperlip, returning all structures in this lineage,

Figure 7. FaceBase search interface. Clicking on one of the categories in the top row (e.g. Anatomy) brings up a set of terms from theOCDM. The user has checked one of these terms (Medial nasal process), which retrieves the three mouse datasets annotated with this term.


including the right and left medial nasalprocess. This list of precursor structures isthen used in a second query that searchesthe Hub database for those datasetsannotated with an Anatomy term thatis one of these precursor structures. Theresult, which is only available in rawform (and hence not shown here) since itis not yet integrated with the Hub searchinterfaces, is the set of data annotatedwith any precursor structures to astructure involved in cleft lip, withoutrequiring the user to know what thosestructures are.

Although the knowledge embodiedin this example is straightforward andlikely already known by most craniofa-cial researchers, this query neverthelessdemonstrates the potential of combiningontological knowledge with queriesover annotated data. As the knowledgein the OCDM becomes more detailed,and as relevant data become moreprecisely annotated, this potential shouldbecome increasingly useful for findingdata relevant to specific phenotypes, not

only from the Hub, but also from otherweb accessible craniofacial data sources.

OCDM Viewer

The queries presented in the last sectionare not ideal for end users since thequeries and their results are onlyavailable in raw form. However, thesequeries can be accessed by applicationsthat present a clickable user interface,thereby hiding the details of the queryfrom the user. As one example of such anapplication, a simple web‐based viewer(available at http://purl.org/sig/ocdm/viewer) displays content from theOCDM. The viewer has been devel-oped for two purposes: (1) it exposes thestructure and content of the OCDM tomembers of the craniofacial researchcommunity as the OCDM is beingdeveloped, because feedback will benecessary to ensure that the OCDMaccurately and completely representsconcepts in craniofacial research and inclinical settings; and (2) it serves as a

demonstration of how a user interfacecan retrieve content of the OCDMusing saved queries in the QueryIntegrator. In fact Figures 3 and 4 arescreenshots of this viewer in action. Eachtime a user clicks a structure a new queryis sent to the QI, with that structure asthe argument. The viewer currentlydisplays class hierarchies and part hierar-chies for several components of theOCDM, as well as mappings of theCHMMO.

Although the current OCDMviewer is designed for ontology authorsrather than for end‐user tasks likeannotation or search, it neverthelessdemonstrates the potential to buildusable applications over the underlyingontology. In our current work we areexamining additional user interfacerequirements in order to add suchfeatures as selection from a developmen-tal timeline, graphs depicting develop-mental relations, and many others. Ournext steps will be to design the userinterfaces for these features, enter any


new required knowledge in theOCDM, and write additional queriesover the OCDM and/or data availableon the web.

DISCUSSION

In this report we have introduced theOntology of Craniofacial Developmentand Malformation, which includes bothhuman and mouse adult and develop-mental anatomy, as well as mappingsbetween them. The use of modelorganisms to enhance understanding ofgenetic determinants of human develop-mental processes is the essence oftranslational medicine. Although geneticexperiments in developing mice provide

The use of model organismsto enhance understanding ofgenetic determinants ofhuman developmental

processes is the essence oftranslational medicine.

detailed insight into the regulatoryprocesses that guide craniofacial devel-opment, these models may not alwayspresent with the anticipated phenotypeor it may be difficult to appreciate orpredict how the homologous structureswill be affected because of subtle differ-ences in development between species.Understanding these differences not onlyaids in recognition of more appropriateanimal models but also helps to betterdecipher the translational impact ofgenetic and epigenetic factors on facialdevelopment and susceptibility to mal-formation. Detailed descriptions of cor-respondences between anatomic regionsthroughout development are needed tolocalize genetic activity with enoughprecision to create meaningful corre-spondences between species. The needfor ontological precision similarly ex-tends to descriptions ofmalformations, asterms that are imprecise regarding thelocation, nature, and timing of a malfor-

mationmight lead to incorrect inferencesregarding the fitness of various animalmodels for specific human diseases. Thegoal of the OCDM is therefore toprovide fine‐grained ontological modelsof craniofacial anatomy, development,and malformation, in support of high‐precision translational craniofacialresearch.

The modular design of the OCDMfollows the tradition of various “slim”

ontology approaches [Courtot et al.,2011; Gene Ontology Consortium,2013], providing applications or otherontologies with the option of usingsome or all of the OCDM componentsas necessary. This approach will supportthe evolution of the OCDM, includingplanned components for modelingmouse malformations (CMMO—theCraniofacial Mouse Malformation On-tology) and mappings between humanand mouse malformations (CHM3OCraniofacial Human‐Mouse Malforma-tion Mappings Ontology). It will alsosupport interoperability of the OCDMwith related ontologies like EMAP[Baldock et al., 1999], especially asmore ontologies move toward thecommon Web Ontology Language(OWL) [World Wide Web Consortium,2010], which is designed to allowdifferent ontologies to interoperate in aworldwide web of knowledge (thesemantic web). In addition, the avail-ability of the OCDM via a web servicemakes it readily accessible to anyapplication, such as theQuery Integratorand the ontology browser describedearlier, or to other groups who wouldlike to use the OCDM to annotate theirown data.

In our future work we will takeadvantage of these capabilities to imple-ment applications for craniofacial re-searchers, who will not need tounderstand the underlying ontology.Example applications include moreintuitive ontology viewers, tools forannotating data, and “intelligent” searchinterfaces for finding data that arecurrently only related by knowledge inthe user’s head. As one example, the usecase illustrated in Table II, findingrelevant mouse gene expression datafor Apert or other syndromes, could be

implemented as a generalized searchinterface that allows a user to select anysyndrome, not just Apert, and have thesystem return relevant mouse geneexpression data for that syndrome. Orthe user could select a set of syndromesand have the system return expressiondata that are relevant for all members ofthe set.

Such intelligent search capabilities,initially over the FaceBase Hub, but laterover other web accessible craniofacialdata sources, should make it much easierthan it is now for both clinicians andresearchers to find candidate genes andother factors that might be involved in aparticular malformation. This type ofretrieved information could in turnsuggest further experiments, the resultsof which could then be made availablethrough the FaceBase Hub or throughother web accessible repositories, forlater intelligent searches. If the resultsof these experiments demonstrate newrelations between genes and pheno-types, and if these results are verified byothers, then this new information couldbe entered into the OCDM, likely withprobabilities to indicate the degree ofconfidence in the relations as is currentlydone with gene annotation. TheOCDM and related ontologies couldtherefore gradually become repositoriesfor the knowledge gained by endeavorssuch as FaceBase.

With our growing recognition ofthe complexities involved in normal andabnormal development and malforma-tions of the human organism it isbecoming increasingly clear to manythat no single human mind is capable ofgrasping all this information at once.Thus, “computable knowledge,” of thetype represented in the OCDM andother ontologies, will become increas-ingly essential, not only for integratingdata, but also for representing thescientific theories that result from analyz-ing these integrated data [Karp, 2001].

REFERENCES

Allanson JE, Biesecker LG, Carey JC, HennekamRCM. 2009. Elements of morphology:Introduction. Am J Med Genet Part A149A:2–5.


Armit C, Venkataraman S, Richardson L, Steven-son P, Moss J, Graham L, Ross A, Yang Y,Burton N, Rao J, Hill B, Rannie D, WicksM, Davidson D, Baldock R. 2012. eMou-seAtlas, EMAGE, and the spatial dimensionof the transcriptome. Mamm Genome 23:514–524.

Ashburner M, Ball CA, Blake JA, Botstein D,Butler H, Cherry JM, Davis AP, Dolinski K,Dwight SS, Eppig JT, Harris MA, Hill DP,Issel‐Tarver L, Kasarskis A, Lewis S, MateseJC, Richardson JE, Ringwald M, RubinGM, SherlockG. 2000. Gene ontology: Toolfor the unification of biology. The GeneOntology Consortium. Nat Genet 25:25–29.

Baldock RA, Dubreuil C, Hill B, Davidson D.1999. The Edinburgh mouse atlas: Basicstructure and informatics. In: Letovsky SI,editors. Bioinformatics databases and sys-tems. Kluwer Academic Press. pp 102–115.

Bard J. 2012. A new ontology (structured hierar-chy) of human developmental anatomy forthe first 7 weeks (Carnegie stages 1–20). JAnatomy 221:406–416.

Bauer S, Kohler S, Schulz MH, Robinson PN.2012. Bayesian ontology querying for accu-rate and noise‐tolerant semantic searches.Bioinformatics 28:2502–2508.

Berners‐Lee T, Hendler J, Lassila O. 2001. Thesemantic web. Sci Am 284:34–43.

Brinkley JF, Detwiler LT. 2012. A query integratorand manager for the Query Web. J BiomedInform 45:975–991.

Brugmann SA, Tapadia MD, Helms JA. 2006. Themolecular origins of species‐specific facialpattern. Curr Top Dev Biol 73:1–42.

Chen C, Mungall CJ, Gkoutos GV, Doelken SC,Köhler S, Ruef BJ, Smith C, Westerfield M,Robinson PN, Lewis SE, Schofield PN,Smedley D. 2012. MouseFinder: Candidatedisease genes from mouse phenotype data.Hum Mutat 33:858–866.

Courtot M, Gibson F, Lister A, Malone J, SchoberD, Brinkman R, Ruttenberg A. 2011.MIREOT: The minimum information toreference an external ontology term. ApplOntol 6:23–33.

Dahdul WM, Balhoff JP, Blackburn DC, DiehlAD, Haendel MA, Hall BK, Lapp H,Lundberg JG, Mungall CJ, Ringwald M,Segerdell E, Van Slyke CE, Vickaryous MK,Westerfield M, Mabee PM. 2012. A unifiedanatomy ontology of the vertebrate skeletalsystem. PLoS ONE 7:e51070.

Depew MJ, Compagnucci C. 2008. Tweaking thehinge and caps: Testing a model of theorganization of jaws. J Exp Zool B Mol DevEvol 310:315–335.

Doelken SC, Kohler S, Mungall CJ, Gkoutos GV,Ruef BJ, Smith C, Smedley D, Bauer S,Klopocki E, Schofield PN, Westerfield M,Robinson PN, Lewis SE. 2012. Phenotypicoverlap in the contribution of individualgenes to CNV pathogenicity revealed bycross‐species computational analysis of sin-gle‐gene mutations in humans, mice andzebrafish. Dis Model Mech 6:358–372.

Du P, Feng G, Flatow J, Song J, Holko M, KibbeWA, Lin SM. 2009. From disease ontology todisease‐ontology lite: Statistical methods toadapt a general‐purpose ontology for the testof gene‐ontology associations. Bioinformat-ics 25:i63–i68.

Fish JL, Villmoare B, Köbernick K, CompagnucciC, Britanova O, Tarabykin V, Depew MJ.2011. Satb2, modularity, and the evolvabilityof the vertebrate jaw. Evol Dev 13:549–564.

Gene Ontology Consortium. 2013. GO Slim andSubset Guide.

Gennari JH, Musen MA, Fergerson RW, GrossoWE, CrubezyM, Eriksson H, NoyNF, Tu S.2003. The evolution of Protégé: An envi-ronment for knowledge‐based systems de-velopment. Int J Hum Comput Stud 58:89–123.

Gray H. 1918. Anatomy of the human body.Philadelphia, Pennsylvania: Lea & Febiger.

Grenon P, Smith B, Goldberg L. 2004. Biodynamicontology: Applying BFO in the biomedicaldomain. Stud Health Technol Inform 102:20–38.

Hayamizu TF, Mangan M, Corradi JP, Kadin JA,Ringwald M. 2005. The Adult MouseAnatomical Dictionary: A tool for annotat-ing and integrating data. Genome Biol 6:R29.

Hayamizu TF, de Coronado S, Fragoso G, SioutosN, Kadin JA, Ringwald M. 2012. Themouse‐human anatomy ontology mappingproject. Database (Oxford) 2012:bar066.

Hochheiser H, Aronow BJ, Artinger K, Beaty TH,Brinkley JF, Chai Y, Clouthier D, Cunning-hamML, DixonM, Donahue LR, Fraser SE,Iwata J, Marazita ML, Murray JC, Murray S,Postlethwait J, Potter S, Shapiro LG, SpritzR, Visel A, Weinberg SM, Trainor PA. 2011.The FaceBase Consortium: A comprehen-sive program to facilitate craniofacial re-search. Dev Biol 355:175–182.

Huss JW III, Lindenbaum P, Martone M, RobertsD, Pizarro A, Valafar F, Hogenesch JB, Su AI.2010. The Gene Wiki: Community intelli-gence applied to human gene annotation.Nucleic Acids Res 38:D633–D639.

Kanzaki S, Takada Y, Niida S, Takeda Y, UdagawaN, Ogawa K, Nango N,Momose A, MatsuoK. 2011. Impaired vibration of auditoryossicles in osteopetrotic mice. Am J Pathol178:1270–1278.

Karp P. 2001. Pathway databases: A case study incomputational symbolic theories. Science293:2040–2044.

Kohler S, Doelken SC, Ruef BJ, Bauer S,Washington N, Westerfield M, Gkoutos G,Schofield P, Smedley D, Lewis SE, RobinsonPN, Mungall CJ. 2013. Construction andaccessibility of a cross‐species phenotypeontology along with gene annotations forbiomedical research. F1000Research http://f1000research.com/articles/2‐30/v1 2:30(doi:10.12688/f1000research.2‐30.v1).

Kundu S, Itkin M, Gervais DA, KrishnamurthyVN, Wallace MJ, Cardella JF, Rubin DL,Langlotz CP. 2009. The IR Radlex Project:An interventional radiology lexicon—Acollaborative project of the RadiologicalSociety of North America and the Societyof Interventional Radiology. J Vasc IntervRadiol 20:S275–S277.

Laue K, Pogoda H‐M, Daniel PB, van HaeringenA, Alanay Y, von Ameln S, Rachwalski M,Morgan T, Gray MJ, Breuning MH, SawyerGM, Sutherland‐Smith AJ, Nikkels PG,Kubisch C, Bloch W, Wollnik B, Ham-merschmidt M, Robertson SP. 2011. Cra-niosynostosis and multiple skeletal anomaliesin humans and zebrafish result from a defect

in the localized degradation of retinoic acid.Am J Hum Genet 89:595–606.

Marazita ML, Mooney MP. 2004. Current con-cepts in the embryology and genetics ofcleft lip and cleft palate. Clin Plast Surg 31:125–140.

McIlraith SA, Son TC, Honglei Z. 2001. Semanticweb services. IEEE Intelligent Syst 16:46–53.

Mungall CJ, Gkoutos GV, Smith CL, HaendelMA, Lewis SE, Ashburner M. 2010. Inte-grating phenotype ontologies across multiplespecies. Genome Biol 11:R2.

Mungall CJ, Torniai C, Gkoutos GV, Lewis SE,Haendel MA. 2012. Uberon, an integrativemulti‐species anatomy ontology. GenomeBiol 13:R5.

Musen MA, Noy NF, Shah NH, Whetzel PL,Chute CG, Story M‐A, Smith B. 2012. TheNational Center for Biomedical Ontology.J Am Med Inform Assoc 19:190–195.

NCHS. 2013. ICD—ICD‐10‐CM—Internation-al Classification of Diseases, Tenth Revision.http://www.cdc.gov/nchs/icd/icd10cm.htm

Niknejad A, Comte A, Parmentier G, Roux J,Bastian FB, Robinson‐Rechavi M. 2012.vHOG, a multispecies vertebrate ontology ofhomologous organs groups. Bioinformatics28:1017–1020.

Olson J, Hofer E, BoxN, ZimmermanA,OlsonG,Cooney D. 2008. A theory of remotescientific collaboration. In: Olson GM,Zimmerman A, Bos N, editors. Scientificcollaboration on the Internet. Cambridge,MA: MIT Press. http://www.ics.uci.edu/�jsolson/publications/CollaborationsInAcademia/TORSC.pdf

Online Mendelian Inheritance in Man. 2010.Apert Syndrome, MIM ID # 101200.

Robin NH, Falk MJ, Haldeman‐EnglertCR. 1993. FGFR‐related craniosynostosissyndromes. In: Pagon RA, Bird TD,Dolan CR, Stephens K, Adam MP, editors.GeneReviews™. Seattle, WA: University ofWashington.

Robinson PN, Mundlos S. 2010. The humanphenotype ontology. Clin Genet 77:525–534.

Rosse C, Mejino JLV. 2003. A reference ontologyfor bioinformatics: The Foundational Modelof Anatomy. J Biomed Inform 36:478–500.

Rosse C, Mejino JLV. 2007. The foundationalmodel of anatomy ontology. In: Burger A,Davidson D, Baldock R, editors. Anatomyontologies for bioinformatics: Principles andpractice. Springer. pp 59–117.

Rosse C, Kumar A, Mejino JLV, Cook DL,Detwiler LT, Smith B. 2005. A strategyfor improving and integrating biomedicalontologies. In: Proceedings, AMIA FallSymposium. Washington, DC. pp 639–643.

Shah NH, Cole T, Musen MA. 2012. Chapter 9:Analyses using disease ontologies. PLoSComput Biol 8:e1002827.

Smith CL, Eppig JT. 2012. The MammalianPhenotype Ontology as a unifying standardfor experimental and high‐throughput phe-notyping data. Mamm Genome 23:653–668.

Sowa JF. 1999. Knowledge representation: Logical,philosophical, and computational founda-tions. Pacific Grove, CA: Brooks Cole.

Spackman KA, Campbell KE. 1998. Com-positional concept representation using


SNOMED: Towards further convergence ofclinical terminologies. Proc AMIA Symp740–744.

Travillian R, Malone J, Pang C, Hancock J,Holand P, Schofield P, Parkinson H. 2011a.The Vertebrate Bridging Ontology (VBO).Bio‐Ontologies 2001.http://bio‐ontologies.knowledgeblog.org/231

TravillianRS, DiatchkaK, Judge TK,WilamowskaK, Shapiro LG. 2011b. An ontology‐basedcomparative anatomy information system.Artif Intell Med 51:1–15.

Washington NL, Haendel MA, Mungall CJ,Ashburner M, Westerfield M, Lewis SE.2009. Linking human diseases to animalmodels using ontology‐based phenotypeannotation. PLoS Biol 7:e1000247.

Whetzel PL, Noy NF, Shah NH, Alexander PR,Nyulas C, Tudorache T, Musen MA. 2011.BioPortal: Enhanced functionality via newWeb services from the National Center forBiomedical Ontology to access and useontologies in software applications. NucleicAcids Res 39:W541–W545.

Whitmore I. 1999. Terminologia anatomica: Newterminology for the new anatomist. AnatRec 257:50–53.

Wilkie AO, Slaney SF, Oldridge M, PooleMD, Ashworth GJ, Hockley AD, HaywardRD, David DJ, Pulleyn LJ, Rutland P,et al. 1995. Apert syndrome results fromlocalized mutations of FGFR2 and is allelicwith Crouzon syndrome. Nat Genet 9:165–172.

World Wide Web Consortium. 2010. The OWLWeb Ontology Language.

Documents

The Ontology of Craniofacial Development and Malformation