|
|
AMIA Annu Symp Proc. 2005; 2005: 639–643. | PMCID: PMC1560467 |
Copyright This is an Open Access article: verbatim copying and redistribution of
this article are permitted in all media for any purpose A Strategy for Improving and Integrating Biomedical Ontologies Cornelius Rosse, MD, DSc,ab Anand Kumar, MD, PhD,d Jose LV Mejino, Jr, MD,a Daniel L Cook, MD, PhD,ac Landon T Detwiler,a and Barry Smith, PhDde a Structural Informatics Group, Departments of Biological Structure, b Medical Education and Biomedical Informatics, and c Physiology and Biophysics, d University of Washington and IFOMIS, University of Saarland, Saarbrücken, Germany, e Department of Philosophy, SUNY at Buffalo The integration of biomedical terminologies is indispensable to the process
of information integration. When terminologies are linked merely
through the alignment of their leaf terms, however, differences in context
and ontological structure are ignored. Making use of the SNAP and
SPAN ontologies, we show how three reference domain ontologies can be
integrated at a higher level, through what we shall call the OBR framework (for: Ontology
of Biomedical Reality). OBR is designed to facilitate
inference across the boundaries of domain ontologies in anatomy, physiology
and pathology. Keywords: ontology integration, top-level ontology, domain ontology, terminology, biomedicine Introduction and Background Ontology, a branch of philosophy, is the theory of what exists in all areas
of reality. The term has of late acquired currency also in the biomedical
domain, where a number of terminologies, developed primarily
for encoding or annotating biological or clinical data, are now commonly
referred to as ontologies. Unfortunately, most such terminologies are
not supported by theories, and they fall short to varying degrees of
conforming to sound ontological practice. 1,2 Even those computational representations of biological entities that were
designated as ontologies from their inception exhibit similar shortcomings. 3Ontologies in biological and medical domains may be sorted into three categories: - Formal, top-level ontologies, such as DOLCE4 and Basic Formal Ontology (BFO)5, which provide domain-independent theories largely through a frame-work
of axioms and definitions, involving categories such as continuant, process and boundary and relations such as is_a (for subtype) and part_of. They are marked by a high degree of representational adequacy and are
designed primarily to be used as controls on the remaining two types of
ontologies.
- Domain reference ontologies, such as the Foundational Model of Anatomy (FMA),6 which declare a theory about a particular domain of reality and make use
of the machinery of formal ontology for sorting and interrelating the
entities that exist in this domain. Reference ontologies are general-purpose
resources designed to generalize to other domains (e.g., from
the human to other species) and also to support a range of different
types of research and clinical applications.
- Terminology-based application ontologies, which are systems of terms (or ‘controlled
vocabularies’) purpose-built and designed
to meet particular needs, such as annotating biological databases (e.g., the
Gene Ontology and other OBO ontologies7) or the medical record (e.g., ICD-10, SNOMED).
Biomedical ontologies are being developed in ever growing numbers; but
unfortunately there is still too little attention paid by the various
separate groups involved to results already obtained by other groups working
in neighboring or even overlapping fields. Thus although several
anatomy terminologies exist already for both human and murine species, new
anatomy terminologies were recently developed for each of these
species in the NCI Thesaurus 8 without any apparent reference to this existing work. The need for ontology alignment and integration is however widely accepted. The
attempts to realize this goal have been limited thus far to the
creation of mere mappings (often at the purely syntactical level) horizontally
between one terminology-based ontology and another. Since
many existing ontologies are of poor quality, however, the provision of
such mappings represents no scientific advance. To achieve the necessary
improvements, we advocate vertical integration between ontologies
in the three categories, of a sort designed also to achieve improvements
in quality of the aligned terminology-based application ontologies. The
type 1 formal, toplevel ontologies should provide the validated
framework for type 2 (reference) ontologies that represent the domains
of reality studied by the basic biomedical sciences. The latter should
then in turn provide the scientifically tested framework for a variety
of type 3 (terminology-based) ontologies developed for specific application
purposes. The basic sciences of anatomy, physiology, pathology, microbiology, genome
science and molecular and developmental biology have served for many
generations as the foundations of clinical medicine. The results of
these basic sciences are applied in all application domains of health
care and biomedical research. Developers of application ontologies intended
to support activities in biomedical research, education and health
care should now similarly be in a position to reuse these results
and to this end reference ontologies for each of the corresponding basic
science domains are urgently needed. At the same time an approach to
ontology development based on vertical correlations between the three
types of ontologies, by providing for the inheritance of sound ontology
construction principles, should promote horizontal interoperability
between different ontologies of types 2 and 3 in ways which will support
new types of automatic inference across the different domains of
biomedicine. In this paper we describe an example for this new paradigm of ontology
development. We report on the process of integrating a domain reference
ontology, exemplified by the FMA, into the framework of BFO, a formal, top-level
ontology, and illustrate the benefits that can be realized
through such integration. The following two sections introduce BFO and
the FMA and related evolving ontologies. We then propose as a hypothesis
the first iteration of an Ontology of Biomedical Reality (OBR). BFO is a formal, top-level ontology based on tested principles for ontology
construction, which subdivides reality into two orthogonal categories 5. The SPAN ontology takes account of occurrents, process-sual entities (events, actions, procedures, happenings) which
unfold over a span of time from their beginning to their ending. The
complementary SNAP ontology ranges over continuants, the participants in such processes, which are entities that endure during
the period of their existence. We can think of the SNAP ontology
as a snapshot of the continuant entities existing at some given instant
of time. The SPAN ontology, in contrast, surveys the processes unfolding
over a given interval of time. Anatomy is a science that studies
biological continuants; physiology studies biological occurrents. Pathology, on
the other hand, is concerned with structural alterations of
biological continuants and with perturbations of biological occurrents
which together are ultimately manifested as diseases. Since processes
cannot exist without their participants, occurrents are entities that depend on corresponding continuants. Entities can thus be sorted also into independent and dependent categories. Cells and organs are independent continuants; a cell surface
or a state of an organism (e.g. of being alive) is a dependent continuant. In
addition to the orthogonal SNAP and SPAN categories, BFO draws
distinctions also between instances (individuals, tokens, particulars) and universals (categories, types, kinds, classes), and furnishes formal definitions
for its high-level categories, and for the relations which link them. 9–11FMA and Associated Ontologies The Foundational Model of Anatomy was initially intended as a terminology
to enhance the anatomical content of UMLS, but in the course of its
development was gradually transformed into a reference ontology for anatomy. 12 Independent evaluations indicate that in its current state the FMA satisfies
a comprehensive suite of requirements deemed to be fundamental
for a sound ontological representation of a life science domain. 3,13The FMA comprehends the structure of the idealized human body – it
deals with canonical anatomy – and ranges over those classes
of anatomical entities (anatomical universals) which exist in reality
through their instances. The root of the FMA’s anatomy taxonomy (AT) is
anatomical entity and its dominant class is anatomical structure. Anatomical
structure is defined as a material entity which has
its own inherent 3D shape and which has been generated by the coordinated
expression of the organism’s own structural genes. This
class includes material objects that range in size and complexity from
biological macromolecules to whole organisms. However, the definition
excludes tumors and other pathological lesions as well as those foreign
organisms and their parts, as well as non-biological entities, which
are introduced into the organism. Portions of body substance (e.g., a
portion of cytosol, intercellular matrix, lymph) and immaterial entities (spaces, surfaces, lines and points) are represented in the FMA in
terms of their relation to anatomical structures. More than two million
instantiations of over 150 relations interrelate AT’s more
than 72,000 classes in several relational networks. 12,14 Consistent with its foundational nature, the FMA is providing a template
for two evolving biomedical domain ontologies: the Physiology Reference
Ontology (PRO) 15 and the Pathology Reference Ontology (PathRO). Ontology of Biomedical Reality To demonstrate how vertical integration with a sound top-level ontology
can support horizontal integration of type 2 ontologies we here present
OBR, a federation of interdependent ontologies which range over the
domains of biomedical reality traditionally studied by the basic biomedical
sciences (). It is rooted in the principles embodied in BFO and the FMA and also
amalgamates the evolving PRO and PathRO and is, like all of these, purpose-neutral. The root of OBR is the universal biological entity, which is primitive (thus
not defined). A distinction is then drawn between the classes biological
continuant and biological occurrent, the definitions of which
are inherited from BFO, whose SNAP-SPAN framework incorporates a methodology
for reasoning simultaneously with both continuants and occurrents. 5 Next we distinguish in both the continuant and occurrent categories classes
of entities that range over single organisms and their parts and
associated processes (instances of organismal entity) from those that
range over aggregates of organisms and over processes in which multiple
organisms participate (instances of extra-organismal entity). The FMA, PRO
and PathRO comprehend organismal entities only. Independent organismal continuants We first subdivide the class organismal continuant into independent and
dependent subcategories. Extrapolating from the FMA’s principles, independent
organismal continuants have mass and are material, whereas
dependent continuants, which are immaterial, do not have mass. Relying
on the FMA’s definition of anatomical structure, we distinguish
anatomical (normal) from pathological (abnormal) material entities; the
latter resulting from processes other than those governed by
the organism’s structural genes. We also extend the FMA’s
representation of material entities by distinguishing structures
from portions of body substances on the basis of the possession by the
former of their own inherent 3D shape. The representation of blood, bile, etc., in
terms of portions caters to the need to give an account
of those processes in PRO and PathRO which involve as participants, for
example, different portions of blood, inspired air, pus, exudate or
cytoplasm. Within the class anatomical structure we make a distinction between canonical
anatomical structures, which exist in the idealized organism, and
variant anatomical structures, which result from an altered expression
pattern of normal structural genes without health related consequences
for the organism (e.g., presence of a middle lobe of left lung, accessory
spleen, or dextrocardia). In contrast, most instances of pathological
structure come into being through the transformation of normal
anatomical structures by processes other than the expression of the
organism’s normal complement of genes (e.g., abscesses, tumors, granulomas, etc.). We retain the traditional paradigm of sorting pathological structures into
categories on the basis of etiology and the pathogenic processes that
generate them (e.g., congenitally abnormal structure, neoplasm, inflammatory
structure, and so on). Since tumors or portions of pus do not exist in the domain of entities
represented in canonical anatomy, pathological structure and portion of
pathological body substance are not subordinated in OBR to the class
material anatomical entity, but rather to its sibling class material
pathological entity. Dependent organismal continuants In analogy with their normal anatomical counterparts, we recognize also
pathological spaces and boundaries such as surfaces and lines. These
immaterial anatomical and pathological entities are dependent continuants, since
their existence depends on corresponding independent continuant
entities. Thus without the stomach, the surface and cavity of the
stomach cannot exist, and the same dependence prevails for the cavity
of an abscess on the abscess or for the surface of a tumor on the tumor. Examples of different classes of dependent continuants are the functions, states
and roles in the domains of physiology and pathology. Functions
are certain sorts of potentials of independent anatomical continuants
for engagement and participation in one or more processes through
which the potential becomes realized. Whereas processes unfold in time, the
function (e.g., the potential of a cell to synthesize a particular
protein) is a continuant, since it, too, endures through time and it
exists even during those times when it is not being realized. The function
to secrete, though dependant on the cell, is independent of the
actual processes (of functioning) that realize this function. Whether or not a function becomes realized depends on the physiological
or pathological state of the associated independent anatomical continuant. By
physiological and pathological state we mean a certain enduring
constellation of values of an independent continuant’s aggregate
physical properties. These physical properties are represented in
the Ontology of Physical Attributes (OPA), which provides the values
for the physical properties of organismal continuants. The states of
these continuants can be specified in terms of specific ranges of attribute
values and a change in these values will become manifest as an observable
and measurable process. The independent continuants that participate in a physiological or pathological
process may play different roles in the process (e.g. as agent, co-factor, catalyst, etc.). Such a process may transform one state
into another; for example a physiological into another physiological, or
into a pathological state. One pathological state is transformed into
another as a pathological structure or disease develops. Organismal Occurrents Paralleling the classification of independent organismal continuants, we
distinguish among organismal occurrents between physiological and pathological
processes (). The transformations of one physiological state into another are instances
of physiological process, whereas processes that transform a physiological
into a pathological state, or one pathological state into another, are
instances of pathological process. The relative balance of
the two kinds of processes as they evolve over time results either in
the maintenance of health or in the pathogenesis of material pathological
entities and thus in the establishment and progression of diseases. Transformation
of a pathological state into a physiological one, manifest
as healing or recovery from a disease, comes about through physiological
processes that successfully compete with and ultimately replace
pathological processes. Function is restored. Processes are extended not only in time but also in space by virtue of
the nature of their participants; their location is the location of their
participants. The synthesis of insulin, for example, takes place within
a beta cell because the protein synthetic apparatus with the potential (function) to
engage in this synthesizing process is included in
the beta cell as part. Conclusions and Discussion We propose the Ontology of Biomedical Reality as a high-level framework
for integrating in robust fashion those more narrowly tailored ontologies
developed to meet the needs which arise in specific domains of life
science and health care. Ontology integration through OBR complements
current reform efforts, for example in the framework of the OBO (Open
Biomedical Ontologies) consortium to bring about a greater degree of
formal rigor in the definition of biomedical relations. 10 We hope that it will greatly improve on the prevailing practice of mapping
between the leaves of structured vocabularies. The benefits of integration
through OBR are entailed by the sound ontological structure
of BFO and of the FMA, which the OBR framework imposes on existing and
emerging ontologies in disparate domains, and this in turn lays the ground
for the drawing of inferences across the boundaries of domain ontologies. OBR integrates the high-level classes of three domain reference ontologies, which
take account of normal and perturbed biological processes along
with the material and immaterial entities that participate in these
processes, encompassing the entire granular spectrum from biological
macromolecules to the whole organism. The integration calls for subordinating
selected classes of reference domain ontologies to corresponding
classes of OBR. For example, the FMA classes anatomical structure
and portion of canonical body substance are subordinated to the OBR class
independent organismal continuant, which also subsumes the PathRO
classes pathological structure and portion of pathological body substance. The
OBR class dependent organismal continuant subsumes the class
immaterial anatomical entity from the FMA, as well as function and physiological state from PRO and
malfunction and pathological state from PathRO. Only PRO and PathRO contribute
to the OBR class biological occurrent. The need for OBR arose in the course of our work on the Virtual Soldier
Project (VSP), which calls for reasoning about traumatic injuries and
prediction of their outcomes. Such reasoning must bridge the divide between
FMA, PRO and PathRO, which together with OPA, constitute the VSP
knowledge base (VSKB). The latter provides input for the amendment of
mathematical models of physiological functions in order to refine their
ability to simulate the behavior of functional systems and predict
the outcomes of traumatic injuries. The VSP serves as the motivator and
evaluation domain for OBR. Preliminary results attest to the potential
and power of an alignment of ontologies along the lines here proposed, suggesting
that the advantages afforded by OBR will translate into
multiple areas of biomedical research and health care, including the
electronic health record. This work was supported in part by a grant from the DARPA Virtual Soldier
Project, Contract # W81XWH-04-2-0012, the Wolfgang Paul Program
of the Alexander Von Humboldt Foundation and the Project “Forms
of Life” of the Volkswagen Foundation. 1. Ceusters W, Smith B, Kumar A, Dhaen C. Ontology-based Error Detection in SNOMED-CT® Medinfo. 2004;2004:482–6. [PubMed] 2. Ceusters W, Smith B, Goldberg L. A terminological and ontological analysis
of the NCI Thesaurus. Methods of Information in Medicine 2005; in press. 3. Smith B, Köhler J, Kumar A. On the application of formal principles to life science data: A case study
in the Gene Ontology. DILS 2004: Data Integration in the Life Sciences. 2004:124–139. 5. Grenon P, Smith B, Goldberg L. Biodynamic ontology: applying BFO in the
biomedical domain. In DM Pisanelli (ed.), Ontologies in Medicine, Amsterdam:IOS Press, 2004,20–38. 9. Smith B, Rosse C. The role of foundational relations in the alignment of biomedical ontologies. Medinfo. 2004;2004:444–448. [PubMed] 10. Smith B, Ceusters W, Klagges B, Köhler J, Kumar A, Lomax J, Mungall C, Neuhaus F, Rector A, Rosse C. Relations in biomedical ontologies. Genome Biology. 2005;6:R46. [PubMed] 11. Donnelly M, Bittner T, Rosse C. A formal theory for spatial representation and reasoning in biomedical
ontologies. IFOMIS Reprints. 2005:1. 12. Rosse C, Mejino JLV., Jr A reference ontology for biomedical informatics: the Foundational Model
of Anatomy. J Biomed Inform. 2003 Dec;36(6):478–500. [PubMed] 13. Zhang S, Bodenreider O. Law and order: Assessing and enforcing compliance
with ontological modeling principles. Computers in Biology and Medicine 2005: in press. 14. Mejino JLV, Rosse C. Symbolic modeling of structural relationships in the
Foundational Model of Anatomy, in U. Hahn (ed.), Proceedings of KR-MED 2004 (First International Workshop on Formal Biomedical Knowledge Representation), 48–62. 15. Cook DL, Mejino JLV, Rosse C. Evolution of a Foundational Model of Physiology: Symbolic representation
for functional bioinformatics. Medinfo. 2004;2004:336–340. [PubMed]
|
This article has been 