Ontologies for the Study of Neurological Disease
Alexander P. Cox1, Mark Jensen1, William Duncan1, Bianca Weinstock-Guttman3, Kinga
Szigeti3, Alan Ruttenberg2, Barry Smith1 and Alexander D. Diehl3*
1
!"#$%&'"(&)*+),-./*0*#-12)3(.4"%0.&1)$&)56++$/*2)56++$/*2)782)39:)
2
)!"#$%&'"(&)*+);%$/)!.$<(*0&.=)9=."(="02)3(.4"%0.&1)$&)56++$/*)9=-**/)*+)!"(&$/)>"?.=.("2)56++$/*2)782)39:)
3
)!"#$%&'"(&)*+)7"6%*/*<12)3(.4"%0.&1)$&)56++$/*)9=-**/)*+)>"?.=.(")$(?)5.*'"?.=$/)9=."(="02)56++$/*2)782)39:)
)
ABSTRACT
@")-$4")A"<6()B*%C)*()&B*)0"#$%$&")A6&)%"/$&"?)*(&*/*<."0)+*%)&-")
0&6?1) *+) ("6%*/*<.=$/) ?.0"$0"0D) ) E-") +.%0&2) &-") 7"6%*/*<.=$/) !.0"$0")
;(&*/*<1)F7!G2).0).(&"(?"?)&*)#%*4.?")$)0"&)*+)=*(&%*//"?2)/*<.=$//1)=*(H
("=&"?)=/$00"0)&*)?"0=%.A")&-")%$(<")*+)("6%*/*<.=$/)?.0"$0"0)$(?)&-".%)
$00*=.$&"?)0.<(0)$(?)01'#&*'02)$00"00'"(&02)?.$<(*0"02)$(?).(&"%4"(H
&.*(0)&-$&)$%")"(=*6(&"%"?).()&-")=*6%0")*+)=/.(.=$/)#%$=&.="D))7!).0)A6./&)
$0)$()"I&"(0.*()*+)&-");(&*/*<1)+*%)J"("%$/)>"?.=$/)9=."(="0)K)$)-.<-H
/"4"/) =$(?.?$&") ;5;) L*6(?%1) *(&*/*<1) &-$&) #%*4.?"0) $) 0"&) *+) <"("%$/)
=/$00"0)&-$&)=$()A")60"?)&*)?"0=%.A")<"("%$/)$0#"=&0)*+)'"?.=$/)0=."(="D)
7!) .0) A".(<) A6./&) B.&-) =/$00"0) 6&./.M.(<) A*&-) &"I&6$/) $(?) $I.*'$&.M"?)
?"+.(.&.*(0)&-$&)?"0=%.A")$(?)+*%'$/.M")&-")%"/$&.*(0)A"&B""().(0&$(="0)
*+)*&-"%)=/$00"0)B.&-.()&-")*(&*/*<1).&0"/+)$0)B"//)$0)&*)"I&"%($/)*(&*/*H
<."0) 06=-) $0) &-") J"(") ;(&*/*<12) N"//) ;(&*/*<12) ,%*&".() ;(&*/*<12) $(?)
N-"'.=$/)O(&.&."0)*+)5.*/*<.=$/)P(&"%"0&D)P()$??.&.*(2)%"+"%"(="0)&*)0.'.H
/$%)*%)$00*=.$&"?)&"%'0).()"I&"%($/)*(&*/*<."02)4*=$A6/$%."0)$(?)&"%'.H
(*/*<."0)$%").(=/6?"?)B-"()#*00.A/"D)P(.&.$/)B*%C)*()7!).0)+*=60"?)*()
&-")$%"$0)*+):/M-".'"%Q0)$(?)*&-"%)?.0"$0"0)$00*=.$&"?)B.&-)?"'"(&.$2)
'6/&.#/")0=/"%*0.02)$(?)0&%*C")$(?)="%"A%*4$0=6/$%)?.0"$0"D)OI&"(0.*(0)
&*)$??.&.*($/)<%*6#0)*+)("6%*/*<.=$/)?.0"$0"0)$%")#/$(("?D)
E-") 0"=*(?) *(&*/*<12) &-") 7"6%*,01=-*/*<.=$/) E"0&.(<) ;(&*/*<1)
F7,EG2).0).(&"(?"?)&*)#%*4.?")$)0"&)*+)=/$00"0)+*%)&-")$((*&$&.*()*+)("6H
%*#01=-*/*<.=$/)&"0&.(<)?$&$D))E-").(&"(&.*()*+)&-.0)*(&*/*<1).0)&*)$//*B)
+*%)&-").(&"<%$&.*()*+)%"06/&0)+%*')$)4$%."&1)*+)("6%*#01=-*/*<.=$/)&"0&0)
&-$&) $00$1) 0.'./$%) '"$06%"0) *+) =*<(.&.4") +6(=&.*(.(<D) 7"6%*H
#01=-*/*<.=$/) &"0&.(<) .0) $() .'#*%&$(&) =*'#*("(&) .() ?"4"/*#.(<) &-")
=/.(.=$/)#.=&6%")60"?).()&-")?.$<(*0.0)*+)#$&."(&0)B.&-)$)%$(<")*+)("6%*H
/*<.=$/)?.0"$0"02)06=-)$0):/M-".'"%Q0)?.0"$0")$(?)'6/&.#/")0=/"%*0.02)$(?)
+*//*B.(<)0&%*C")*%)&%$6'$&.=)A%$.().(R6%1D))7,E).0)A".(<)?"4"/*#"?)$0)
$()"I&"(0.*()&*)&-");(&*/*<1)+*%)5.*'"?.=$/)P(4"0&.<$&.*(0D)
1
INTRODUCTION
The field of neurology deals with a diverse domain of diseases related to the functioning of the nervous system in all
its aspects, including diseases resulting from disorders of
the central, peripheral, and autonomic nervous systems.
Neurological diseases may exhibit both acute and chronic
courses, affect a variety of cell types and anatomical regions
of the body. They are manifested via a variety of mechanisms, including cell-autonomous disorders, unregulated
protein aggregation, autoimmune conditions, and vascular
pathology, which, depending on the disease, may occur
alone or together in various combinations (Ropper et al.,
2005; Merritt and Rowland, 2000). At a different level of
granularity we see neurological diseases that affect cognitive as well as mental functioning. Following Ceusters and
*
To whom correspondence should be addressed: addiehl@buffalo.edu
Smith (2010), we maintain that mental diseases are (at least
primarily) special kinds of neurological diseases in the sense
that the disorder, which serves as the material basis for the
disease, is a part of an anatomical structure in the organism
responsible for producing and maintaining cognitive representations and behavior. For example, a variety of neurological conditions result in dementia, such as Alzheimer’s and
Parkinson's disease, and many of the late-onset leukodystrophies.
We have recently begun building a new ontology for the
domain of neurological diseases – the Neurological Disease
Ontology (ND). ND is an ongoing project that aims to accurately represent every facet of neurological diseases in as
much detail as possible. This includes their clinical presentation, diagnosis, treatment, physical manifestation, course
of development, genetic and physical bases, and more. ND
is still in the early stages of development, but is rapidly
growing to include more of these facets. While our ultimate
goal in developing ND is to provide a comprehensive account of all neurological diseases, it has three initial areas of
focus: Alzheimer’s disease (AD), multiple sclerosis (MS),
and stroke and cerebrovascular events. At this time, the
most progress has been made on AD and other diseases that
result in dementia, but work is currently under way on representing MS and associated demyelinating diseases as well
as on representing stroke and cerebrovascular disease.
As a corollary to ND, we have begun development of the
NeuroPsychological Testing Ontology (NPT) to represent
neuropsychological assessments such as the Folstein MiniMental State Examination (MMSE), the Trail- Making Test,
the Hopkins Verbal Learning Test, and the Wechsler
Memory Scale. These standardized assessments are useful
for identifying the presence and degree of cognitive impairment in patients (Lezak et al., 2004). An initial goal of
the NPT project is to test hypotheses about the diagnosis of
AD based on the results of neuropsychological assessments.
Part of the development of NPT necessitates reference to
aspects of cognitive functioning. For example, MMSE produces scores that are indicative of impairment in certain
functional cognitive domains such as language, executive
function, or memory. A challenge we have encountered is
how to connect these commonly described cognitive do-
1
Cox et al.
Figure 1: A subset of OGMS and ND and some connections to external ontologies.
mains to functioning on the side of the organism. We see
this as an excellent opportunity to connect ND and NPT
with work in the Mental Functioning Ontology (MF) as well
as with the Mental Disease Ontology (MD). Ideally we hope
to drive development in both. For example, an extension of
MD that represents dementia from the perspective of it being a mental disease or syndrome could then be linked via
logically defined relations to classes in ND.
We plan to build ND over the long-term in a collaborative manner with other groups focused on representing particular neurological diseases as modules within ND. Our
work is intended to be OBO-Foundry compliant and builds
upon the paradigm established by Ontology for General
Medical Sciences (OGMS) for the representation of entities
in the domain of medicine and disease (Scheuermann et al.
2009).
2
METHODS
ND and NPT are being curated using both top-down and
bottom-up approaches to the creation of classes within the
ontology. A major aspect of the top-down approach for ND
has involved analyzing what types of neurological diseases
exist and how they ought to be represented within the ontol-
2
ogy according to their relevant characteristics. Of some
concern is how our strategy will fit with other disease ontologies. A key element includes deciding what other types of
entities should be represented in ND in order to accurately
represent the neurological diseases as well as how the relationships between these classes should be represented. For
instance, the class ‘neurological disease’ currently includes
‘neurodegenerative disease’, ‘infectious neurological disease’, ‘demyelinating disease’, and ‘vascular neurological
disease’ as four of its subclasses. The inclusion of these
subclasses was driven by our decision to focus, as much as
possible, on representing neurological diseases from the
perspective of their etiology. For example, it is part of the
logical definition for ‘neurodegenerative disease’ that all
realizations of these diseases involve some process of neurodegeneration. This top-down approach provides ND with
its primary structure.
Due to the complex nature of neurological diseases, as
well as the diversity of perspectives from which they are
studied and classified, we have also included additional immediate subclasses of ‘neurological disease’. For example,
‘central nervous system disease’ and ‘peripheral nervous
system disease’ are included as subclasses of ‘neurological
Ontologies for the Study of Neurological Disease
disease’. Currently we do not explicitly assert any disease
as a subclasses of these classes, however ND is being built
using axioms that will allow an ontological reasoner to automatically create an inferred hierarchy of neurological disease types based on anatomical structure or genetic basis.
This approach allows ND more versatility without committing it to a single perspective or creating confusion by
switching between perspectives within the asserted hierarchy. Another example of this approach is creation of the
defined class ‘disease resulting in dementia’, which has a
limited number of asserted subclasses, and was created to
provide a reference class from which to allow a reasoner to
infer a hierarchy of all diseases that result in dementia.
While the top-down aspect of the project is essential to
shaping the development of ND, it is the bottom-up aspect
of the project that provides the bulk of the information. In
particular, it is this approach that results in the creation and
refinement of the definitions for terms in ND. We have
consulted primary research articles, review articles, medical
professionals, and other sources to inform the development
of ND. This process has led to the inclusion of new terms in
ND as well as more detailed classifications of particular
neurological diseases. Both approaches are necessary for the
completion of the project.
Development of NPT is based upon analyses of neuropsychological tests to drive the development of classes for
the representation of neurological assays and their results.
Many neuropsychological tests have multiple subtests, and
these are being captured within the ontology as well. Neuropsychological tests assay domains such as verbal and visual-spatial memory, executive function, and linguistic functions. NPT is being developed to allow the integration of
scores from different neuropsychological tests and subtests
so that results for patients who have been tested using different protocols can be queried and grouped appropriately.
ND and NPT are built using Protégé 4.1 as OWL2 ontologies. The importation of classes from other ontologies according to the MIREOT standard has been achieved using
OntoFox (Xiang, 2010).
Both ND and NPT are being developed according to
OBO Foundry principles (Smith et al., 2007) and is being
done in cooperation with the related efforts to develop
ontologies for representing Mental Disease (MD) and
Mental Functioning (MF) (Hastings et al. 2012a and
2012b).
Ontology Name
Basic Formal Ontology (BFO)
Ontology for General Medical Sciences (OGMS)
NIF-Dysfunction and Disease Ontology (DO)
Relation Ontology (RO)
Protein Ontology (PR)
Foundational Model of Anatomy (FMA)
IAO, PATO, ChEBI, GO, CL, and OBI
Figure 2: :)#*%&.*()*+)&-")7!)?.0"$0")-."%$%=-1D
3
RESULTS
The Neurological Disease Ontology is being built according
to OBO Foundry principles as an extension of OGMS,
which provides a set of general reference classes related to
diseases, their patients, and diagnoses (Scheuermann et al.
2009). OGMS follows the paradigm of the Basic Formal
Ontology (BFO). Figure 1 illustrates the layers of granulariUse in ND
Top-level reference ontology
Mid-level reference ontology
Externally referenced disease classes
Imported relation types
Select classes for proteins imported via MIREOT
Select classes for anatomical structures imported via MIREOT
Select classes imported via MIREOT
Table 1. External ontologies used by the Neurological Disease Ontology.
3
Cox et al.
ty captured by the relations between ND, OGMS, and BFO
as well as IAO and OBI. Furthermore, we are ensuring that
ND is compliant with the pre-release revised version of
BFO – BFO 2.0, and the revised version of OGMS that is
also compliant with BFO 2.0.
In building ND, we have relied upon a number of
sources, including reference works, review articles, and other ontologies, such as NIF-Dysfunction and the Disease
Ontology (DO) (Bug et al., 2008; Larson & Martone, 2009).
Based on these sources we have curated a high-level disease
hierarchy that we believe presents a useful initial approach
to categorizing neurological diseases, a section of which is
shown in Figure 2. We go beyond earlier efforts at creating
disease ontologies by providing textual definitions for every
disease class and by incorporating logical definitions in order to relate classes for diseases and other entities in ND to
other classes in ND and to separate ontologies (See Table 1
for a summary).
These high level disease classes provide a framework for
the in depth curation of ND ontology modules intended to
represent neurological diseases in extensive detail. At the
University at Buffalo, our initial efforts are focused upon
the areas of Alzheimer’s disease and other diseases resulting
in dementia, multiple sclerosis, and stroke and cerebrovascular disease. As an early stage ontology development project, ND currently contains approximately 400 classes;
about 250 classes have textual definitions; more than 50
classes have logical definitions; more than 150 classes have
external references; and there are nearly 200 children of the
class ‘disease’. In addition to disease classes, ND has a
heavy focus on diagnosis, syndrome, disorder, and protein
classes among others in order to fully represent all of the
various aspects of neurological diseases.
In building NPT we have relied upon source tests, such
as the Folstein Mini-Mental State Exam, as well as upon
textbooks and articles about particular neuropsychological
tests (Lezak et al., 2004; Mitrushina et al., 2005). NPT is
built using the schema for representing assays that has been
developed in OBI and consequently currently imports all of
OBI. At a later point, we will rely upon a slimmed (MIREOTed) version of OBI. At the moment, there are more than
250 NPT specific classes, but we expect this to grow quickly as we add representations of additional neuropsychological tests. Figure 3 shows a portion of NPT for the representation of the MMSE.
4
DISCUSSION
Our use cases in building these ontologies include annotation of clinical studies in neurology as well as annotation of
patient records. Particularly for the latter case we expect ND
and NPT to complement each other, with ND providing
terms for representing the diagnoses of patients based on
their signs and symptoms, and associated phenotypes. NPT
will provide a very detailed set of classes for annotation of
neuropsychological measures that may be used in the formation of a patient’s clinical picture, which is used to reach
a diagnosis. These diagnostic conclusions are annotated as
Figure 3: A portion of the representation of the MMSE assay in NPT.
4
Ontologies for the Study of Neurological Disease
an instance of a diagnosis class in ND. The diagnosis classes are linked to the disease classes in ND, which themselves link via their logical definitions to other classes in
ND such as the disorder which serves as the material basis
of the disease, and then, in turn, to other ontologies such as
PR.
In developing ND and NPT we recognize the need to coordinate with other ontology development efforts in related
domains. In Ceusters and Smith (2010), for instance, the
framework for what are now named the Mental Functioning
Ontology (MF) and the Mental Disease Ontology (MD) was
presented. Neurological diseases by their very nature often
affect cognitive and mental functioning, for instance in any
disease that results in dementia, such as Alzheimer’s disease, and often lead to mental diseases, such as depression
in MS or epilepsy patients. In developing ND we will need
to ensure representation of conditions such as dementia or
depression are coordinated with MF and MD, such that a
class representing a clinical phenotype of “depression in
conjunction with multiple sclerosis” may have a parent class
of “depression” in MD. Moreover we feel that our work can
aid in a bottom-up approach to developing MD and MF.
Furthermore, we believe our work on NPT will be valuable for the annotation of neuropsychological data not just
for patients with neurological disease, but also for studies of
general mental functioning and in testing in patients with
mental diseases. Thus, our work on NPT will hopefully
prove of value for a number of related domains in addition
to that of neurological diseases, and will eventually be complemented by ontologies for other types of assessments of
nervous system function and anatomy, such as an MRI imaging ontology.
REFERENCES
Bug, W. J., Ascoli, G. A., Grethe, J. S., Gupta, A., Fennema-Notestine, C.,
Laird, A. R., et al. (2008). The NIFSTD and BIRNLex vocabularies:
Building comprehensive ontologies for neuroscience. Neuroinformatics,
6(3), 175-194.
Ceusters, W. and Smith, B. (2010). Foundations for a realist ontology of
mental disease. Journal of Biomedical Semantics, 1(1), 10.
Hastings, J., Smith, B., Ceusters, W., Jensen, M., and Mulligan, K. (2012a).
The
mental
functioning
ontology.
Available
at
http://code.google.com/p/mental-functioningontology/
Hastings, J., Smith, B., Ceusters, W., Jensen, M., and Mulligan, K.
(2012b). Representing mental functioning: Ontologies for mental health
and disease. See proceedings for this workshop: Towards an Ontology
of Mental Functioning, ICBO 2012, Graz, Austria.
Larson, S. D., and Martone, M. E. (2009). Ontologies for Neuroscience:
What are they and What are they Good for? Frontiers in Neuroscience,
3(1), 60-67.
Lezak, M. D., Howieson, D. B., and Loring, D. W. (2004). Neuropsychological assessment (4th ed.). Oxford: Oxford University Press.
Merritt, H. H. and Rowland, L. P. (2000). Merritt's Neurology (10th ed.).
Philadelphia: Lippincott Williams & Wilkins.
Mitrushina, M., Boone, K. B., Razani, J. and D’Elia, L. (2005). Handbook
of Normative Data for Neuropsychological Assessment (2nd ed.). Oxford: Oxford University Press.
Ropper, A. H., Adams, R. D., Victor, M., Brown, R. H., and Victor, M.
(2005). Adams and Victor's Principles of Neurology (8th ed.). New
York: McGraw-Hill Medical Pub. Division.
Scheuermann, R., Ceusters, W., and Smith, B. (2009). Toward an ontological treatment of disease and diagnosis. In AMIA Summit on Translational Bioinformatics, San Francisco, California, March 15-17, 2009,
pages 116–120. Omnipress.
Smith, B., Ashburner, M., Rosse, C., Bard, J., Bug, W., Ceusters, W.,
Goldberg, L. J., Eilbeck, K., Ireland, A., Mungall, C. J., The OBI Consortium, Leontis, N., Rocca-Serra, P., Ruttenberg, A., Sansone, S.-A.,
ACKNOWLEDGEMENTS
We would like to thank Ralph Benedict, Ph.D., of the Department of Neurology, University at Buffalo, for guidance
in understanding neuropsychological testing, and Naveed
Chaudhry, Marcus Ng, and Donat Sule for assistance with
term development in ND.
Scheuermann, R. H., Shah, N., Whetzel, P. L., and Lewis, S. (2007).
The OBO Foundry: coordinated evolution of ontologies to support biomedical data integration. Nat Biotechnol., 25(11), 1251–1255.
Xiang Z, Courtot M, Brinkman RR, Ruttenberg A, and He Y. (2010). OntoFox: web-based support for ontology reuse. BMC Res Notes, 3, 175.
5