|
|
AMIA Annu Symp Proc. 2005; 2005: 734–738. | PMCID: PMC1560542 |
Copyright This is an Open Access article: verbatim copying and redistribution of
this article are permitted in all media for any purpose Of Mice and Men: Design of a Comparative Anatomy Information System Ravensara S. Travillian, MS, MA,1,2 John H. Gennari, PhD,1,2 and Linda G. Shapiro, PhD1,2,3 1 Structural Informatics Group 2 Dept. of Medical Education & Biomedical Informatics 3 Dept. of Computer Science & Engineering; University of Washington, Seattle, WA 98195 In previous work, we proposed an approach called the Structural Difference
Method (SDM) to correlating the anatomy of Homo sapiens with selected
species 1, using the Foundational Model of Anatomy (FMA) 2,3 as a framework and graph matching as a method, for determining similarities
and differences between species. In this paper, we present the design
of a comparative anatomy information system that utilizes the SDM
and allows users to issue queries to determine the similarities and
differences between two species. Our system will serve as a pilot project
for cross-species anatomical information collection, storage, and
retrieval. The underlying data structure of a mapping, and the syntax
and semantics of the system's query language, are presented. The goal of this work is to build a comparative anatomy information system
to which users can pose queries about the similarities and differences
between two species. In our prior work 1 we have developed an approach called the Structural Difference Method (SDM), in
which each species is represented by an attributed graph, and
graph matching is used to determine the similarities and differences. Using
the Foundational Model of Anatomy (FMA) for humans 2,3 as our framework, we have developed a partial mouse anatomy ontology (MAO) that
can be used for comparisons. Our partial ontology differs from existing ontologies in that, based entirely
on structure, it can serve as a reference ontology for mouse anatomy, and
as a means for integrating different ontologies with other
functions 2. In this respect, it differs from other ontologies, such as GALEN or EMAP, whose
design reflects their applications to pathology and/or clinical
practice. In its capacity as a reference ontology, we expect our
application to ultimately be able to serve as a bridge among domain- and
view-specific anatomical applications, such as, for example, integrating
the mouse anatomy ontologies currently under development with the
human anatomy component of large clinical ontologies, such as SNOMED-CT. To compare two species, a mapping between them must be constructed and
represented as a computer data structure. Since both the FMA and the MAO
have been implemented in the Protégé-2000 frame-based
knowledge system, we have designed the mappings between mouse and human
as Protégé classes that can link the two ontologies
and provide a resource for a query system. The information system proposed in this paper will accept queries posed
by the user about similarities and differences in human and mouse anatomy. The
implementation of the pilot version of the comparative anatomy
system will be a single database of mappings, from which the query
engine will access and return a result set. The proposed graphical user interface for the application is shown in . The system is currently under development, so this proposed interface
is a mockup, but serves to illustrate the kinds of queries supported
by the system. Because the user interface is designed to conform to our syntax (described
below), the user is not required to remember the syntax of the different
queries—rather, she can form a query by selecting choices
which serve to specify the form of the query. The Mapping direction, From
structure, To structure, and Query fields are options selected
by the user to specify the information requested from the database, and
the Mapping results box is where the application displays the results
of the processed query. For example, if the user wanted to ask “How
does the human lung differ from the mouse lung?'', she
would select Mapping direction: Human → Mouse, From structure: Lung, To
structure: Lung, Query: differs-from, and then she would
click the Submit button. The results would appear in the Mapping results
box, and would include such information as the human left lung has 2 lobes
while the mouse left lung has only 1, the human right lung has 3 lobes
while the mouse right lung has 4, etc. The Mapping direction list box specifies which species is the source and
which is the target of the mapping, thus permitting unidirectional as
well as bidirectional queries. The + sign by the structure names
indicates that a hierarchy can be expanded, and it is by selecting
a class in the hierarchy that the granularity of the comparison is specified. We
expand the hierarchy initially along the part-of links, rather than is-a, as our experience shows that use of the partonomy is the most intuitive
way for bio-researchers to think about anatomy. Additionally, use of
the partonomy is consistent with the JAX mouse ontology 4. The Query radio buttons allow the user to select a query relationship, which
will use the specified structures as arguments. Once a query relationship
has been specified, the user can proceed by clicking the Submit
button, or she can click the Clear button to reset the selections
and start over again. When a query has been submitted, the mapping result set—i.e., a set of descriptions of similarities and differences for that structure
across those species, as calculated by the SDM—is returned
by the application in the Mapping results box. The results of the query
can be saved by clicking the Save button, or printed by clicking the
Print button. The anatomical mapping data structure and the syntax and semantics of the
system's query language are particularly significant, and will
be discussed in more detail below. Mappings are the data structure at the heart of the proposed information
system. In order to be able to create a mapping of anatomical structures
across species, the structures must be formally represented in a
way that supports the mapping. In earlier work, it has been demonstrated
that the Foundational Model of Anatomy (FMA) describes multiple directed
acyclic graphs (DAG) 5, and so, in order to create the mappings, we develop the appropriate graphs, one
for the human structure, and one for the mouse structure. In
these graphs, the nodes represent anatomical entities, and the edges
represent the structures among those entities. This representation is
consistent with the one used by Hayamizu's Adult Mouse Anatomical
Dictionary 6, as well, so we expect to be able to extend this system to other ontologies. A Mapping, then, is a correspondence between a structure in the source
species and a structure in the target species. As developed in Travillian 2003, there
are two main kinds of mappings: Node mappings and Edge
mappings, corresponding to the components of the directed graph described
by the FMA. The structures which are mapped across species are selected
on the basis of homology (evolutionary relatedness); homoplasy (similarity
of appearance) and analogy (similarity of function) are not
considered in creating mappings. At a conceptual level, a Mapping across Species between Anatomical structures
can be represented as in , which shows Mappings between the human and mouse Prostates at the Organ
level. Note that the graph is a composite of the is-a and has-member graphs; these relationships were selected to emphasize differences, as
the mouse and human prostates differ in some important and non-intuitive
ways which are demonstrated efficiently in the is-a hierarchy. The fact that the mouse has five prostates, where the human
has one, and that homologies between the mouse organs and the parts of
the human organ have not been definitively established, represent important
considerations in comparative medicine. Since different pathologies
or resistance to pathology (prostatic carcinoma, benign prostatic
hyperplasia, or no disease) arise in different parts of the human prostate, and
since those parts and their pathologies correlate to the embryonic
origin of the structures, the importance of establishing those
homologies to draw the Mappings is clear. Our approach, which maps concepts (rather
than terms) on the basis of homology, has the additional
benefit of resolving discrepancies in term usage between the human
and veterinary medical communities: when the same term is used for different
concepts (homonymy), they are not mapped in our system (i.e., the face validity does not trigger a false match between the different
concepts. Similarly, our system correctly handles synonyms, preferred
terms, and deprecated terms as well, correctly matching the concepts
and not missing the match because of the superficial term discrepancy. The edges of the graph in green represent isomorphisms, or anatomical identity: one-to-one, onto, and structure-preserving. For example, the
anatomical abstraction Lobular organ in the mouse is isomorphic to the
Lobular organ in the human. The edges of the graph in blue represent
non-isomorphic matches. For example, there is a 5:1 mapping between the
five different mouse prostate organs and the single human prostate. The
edges in red represent null mappings. For example, there is no corresponding
Set of human prostates to map to the Set of mouse prostates, so
that constitutes a null mapping. A single mapping can answer a query
such as “What is the structure in the human that corresponds
to the liver in the mouse?”. A unidirectional comparison consists of a hierarchy of Mappings. A root
for the mapping, and the depth to which the comparison is to be pursued, are
chosen, and all the mappings for structures in the hierarchy beginning
at the root and proceeding to the chosen depth, make up the unidirectional
comparison. A unidirectional comparison can thus answer
a query such as “What are the structures in the mouse mammary
gland that are missing in the human mammary gland?”. A cross-species, or
bidirectional, comparison consists of two complementary unidirectional
comparisons. To implement this functionality, the underlying Mapping data structure
contains pointers in both directions between species: i.e., the human
can be either the source or the target species, as can the mouse. Both
directions are necessary for a complete answer to queries on similarities
and differences between species, as, from the user's point of
view, the answer returned to the query “what is the difference
between the human and mouse prostates?” should be the same
as the answer returned to the query “what is the difference between
the mouse and human prostates?”. This data structure provides
that consistency of response, yet at the same time allows a more
refined query to return a more granular answer, depending on the level
of detail the user wishes to specify. Although the usual query will
be bidirectional, there will be users who want information in one direction
only. For example, a user may want to know which prostatic lobe
in the human is homologous to the murine dorsal prostate. This structure
will be able to accommodate those queries as well. The FMA is implemented in Protégé-2000 7. Following the suggestion of Bernstein and Pottinger 8 that mappings should be first-class objects, we have implemented mappings
as Protégé classes. Mappings are implemented in Protégé in
the following manner: the Protégé template
slots for Mapping are the two Species being compared (as
in ), and the two corresponding Anatomical structures. Species names are required
to always be single; Anatomical structures can be 1 or more in
a particular Species. Cardinality specifies whether the correspondence
is 1:null, null:1, 1:1, 1:many, many:1, many:many, many:null, or null:many. The slots for Node mapping are inherited verbatim from the Mapping class; Edge
mappings have the additional slot Relationship to describe which
relationship is being compared across Species for the given Anatomical
structures. These examples demonstrate the definitions for the different
kinds of Mappings. A Cross-species comparison is made up of all
the Mappings of the Anatomical structure at the level under comparison. We propose to use these structures to return answers to anatomical queries
about similarities and differences between these structures—the
Mapping contains the information about similarity and differences
of particular discrete structures, and the Cross-species comparison
provides the context (hierarchy) for those structures in relation to other
anatomical structures. For the purpose of defining this comparative anatomy information system, it
is useful to draw a distinction between different kinds of queries, based
on how many models the system handles at a time. These classifications
will specify what types of queries our system handles, and what
is outside its scope. We define the classification of a query as follows: Single-species queries hold for species models taken one at a time. For
example, in the human, the Heart is inside the Thoracic cavity, so the query "what is the relationship between Heart
and Thoracic cavity [ implied: in the human]?'' is a single-species query. Note that a single-species
query can be simple or compound; the classification of the query
refers to the number of species models participating in the query, NOT
to the complexity of the query. Single-species queries currently can
be the basis of queries in the FMA using the Emily graphical user interface 9, and involve existence, location, connectivity, and similar features of
anatomical structures. Two-species queries hold for species models taken two at a time, and are
the basis of what is unique about our proposed system. They involve
comparisons between anatomical structures across two different species, such
as “how is the human prostate different from the mouse
prostate?”. Two-species queries involve similarity, difference, homology, identity, and
synonymy of anatomical structures in two different
species, as described below. While the concepts of homology, identity, and
synonymy overlap to some degree in natural language, the syntax
below suffices to deal with them at the level of the users' needs. Higher-degree
queries represent future work, and will explicitly
not be treated in this specification. We propose to develop the syntax
for two-species queries, as follows. We propose to use this syntax as the basis for queries and responses about
anatomical similarities and differences between the human and the
mouse. This notation represents an abstraction of the basis for the queries
and responses; there will be a low-level syntax that is used by
the system for accessing and returning information, as well as a higher-level
graphical user interface for the users of the system. Semantics While the details remain to be determined, some of the semantics of the
query language are already emerging from the information gathered to
date. Queries will be of two major types, set and Boolean. Set queries
will return result sets, such as the set of shared mappings between two
species for a structure at a given level of granularity. Boolean-type
queries will, for example, return T or F when the user queries whether
structures in two different species map to each other. The semantics
of the proposed operators are as follows. Set queries The set query operators are differs-from, similar-to, shared, not-shared, and
union. - species1.anatomical-entity1 differs-from species2.anatomical-entity2 returns the difference between anatomical-entity1 in
species1 and anatomical-entity2 in species2. If they are isomorphic, it
will return null.
- species1.anatomical-entity1 similar-to species2.anatomical-entity2 returns the complement of the set returned
by (species1.anatomical-entity1 differs-from species2.anatomical-entity2), which is all of the similarities between
them.
- species1 shared species2 returns the set of non-null mappings between anatomical entities
of species1 and those of species2.
- species1 not-shared species2 returns the set of null mappings between anatomical entities
of species1 and those of species2. In other words, it is the inverse operation
of shared.
- species1 union species2 returns the set of all (null as well as non-null) mappings between
anatomical entities of species1 and those of species2.
Boolean queries The Boolean query operators are is-different? and is-homologous?. - species1.anatomical-entity1 is-different? species2.anatomical-entity2 returns T if species1. anatomical-entity1 does
not map to species2.anatomical-entity2, and F if the two anatomical
entities do map to each other.
- species1.anatomical-entity1 is-homologous? species2.anatomical-entity2 returns F if spe-cies1.anatomical-entity1 does
not map to species2.anatomical-entity2, and T if the two anatomical
entities do map to each other. In other words, it is the inverse operation
of is-different?.
These Boolean and set query operators suffice to deal with the questions
of similarity and difference that a user would ask the system about
the comparisons between mouse and human anatomy, and this aim serves to
provide the structure (syntactic and semantic) for those operators. Many contemporary observers 10,11 have remarked upon the increasing need for extrapolating information from
one species to another, which has been highlighted by contemporary
research in bioinformatics, genomics, proteomics, and animal models of
human disease, as well as other fields. The amount of anatomical and
associated medical information emerging from animal modeling in comparative
medicine and comparative genomics is increasing at an exponential
rate, calling for innovative techniques in evaluating, organizing, and
managing that information for researchers and clinicians. In addition, the
increasingly interdisciplinary nature of medical research has
greatly increased the base of users, and the corresponding need, for
such an information system. Therefore, in addition to rigorous attention
to the quality of the anatomical information involved, such a system
must be flexible and extensible enough to accommodate different information
views, depending on the needs of the user—whether a bench
scientist, a clinician, or a student. In order to successfully manage this information, a systematic, principled
way of correlating anatomical information across species is needed. Because
the FMA has the necessary qualities to serve as the basis for
a sound and complete pan-vertebrate metamodel, we base our information
system on FMA models of the human and of the mouse. In this paper, we propose a pilot comparative anatomy information system
to meet this need. Our proposed information system builds on our previous
work in correlating the anatomy of Homo sapiens with selected species, using the Foundational Model of Anatomy (FMA) as
a framework, and graph matching as a method. It will be able to answer
queries regarding cross-species similarities and differences in structural
phenotypes, and it addresses important scientific questions in
both medical informatics and comparative anatomy. In informatics, the problem of ontology alignment has been a promising
research area for decades, yet the inherent complexity of comparing such
different anatomical data at so many levels of resolution for so many
species poses a challenge far greater than the domain of most ontology
alignments, and carries the promise of developing techniques and tools
that can be applied to genomics ontology alignment problems, taken
as another level of anatomical complexity. As well, in comparative anatomy, the
structure and organization of massive amounts of anatomical
data in one resource will serve multiple purposes of making information
accessible and visualizable in different views for different users
with different information needs, as well as for identifying gaps and
inconsistencies in the scientific literature to facilitate future research. In
this way, it is in accord with the mission of the Standards
and Ontologies for Functional Genomics (SOFG) consortium, whose stated
goal is “to bring together biologists, bioinformaticians, and
computer scientists who are developing and using standards and ontologies
with an emphasis on describing high-throughput functional genomics
experiments.” 12 To this purpose, they have developed SOFG Anatomy Entry List (SAEL), an
entry mechanism for computational access to anatomy resources, to facilitate
automated information retrieval by biologists, informaticians, and
software developers 13. We hypothesize that our homology-based concept-mapping system will prove
to be an initial step toward meeting these needs. This work was supported by the University of Washington National Library
of Medicine Informatics Training Grant (1T15LM07441-01). 1. Travillian RS, Rosse C, Shapiro LG. An approach to the anatomical correlation of species through the Foundational
Model of Anatomy. AMIA Annu Symp Proc. 2003:669–73. [PubMed] 2. Rosse C, Mejino JL., Jr A reference ontology for biomedical informatics: the Foundational Model
of Anatomy. J Biomed Inform. 2003 Dec;36(6):478–500. [PubMed] 3. Rosse C, Shapiro LG, and Brinkley JF. The Digital Anatomist Foundational
Model: Principles for defining and structuring its concept domain. Proceedings 1998 American Medical Informatics Association Annual Symposium, November 1998. 5. Mejino JL, Jr, Rosse C.Conceptualization of anatomical spatial entities in the Digital Anatomist
Foundational Model. Proc AMIA Symp. 1999:112–6. 6. Hayamizu TF, Mangan M, Corradi JP, Kadin JA, Ringwald M. The Adult Mouse Anatomical Dictionary: a tool for annotating and integrating
data. Genome Biology. 2005;6:R29. [PubMed] 7. Noy NF, Crubezy M, Fergerson RW, Knublauch H, Tu SW, Vendetti J, Musen MA. Protege-2000: an open-source ontology-development and knowledge-acquisition
environment. AMIA Annu Symp Proc. 2003:953. [PubMed] 8. Bernstein, PA, Levy AY, Pottinger RA. A Vision for Management of Complex
Models. Microsoft Research Technical Report MSR-TR-2000-53, June 2000. 9. Shapiro LG, Chung E, Detwiler LT, Mejino JL, Jr, Agoncillo AV, Brinkley JF, Rosse C. Processes and problems in the formative evaluation of an interface to the
foundational model of anatomy knowledge base. J Am Med Inform Assoc. 2005 Jan–Feb;12(1):35–46. Epub 2004 Oct 18. [PubMed] 10. Linazasoro G. Recent failures of new potential symptomatic treatments for Parkinson's
disease: causes and solutions. Mov Disord. 2004 Jul;19(7):743–54. [PubMed] 11. Noble M, Dietrich J, Noble M, Dietrich J. The complex identity of brain tumors: emerging concerns regarding origin, diversity
and plasticity. Trends Neurosci. 2004 Mar;27(3):148–54. [PubMed]
|
