Objective: To investigate the feasibility of deriving an indirect alignment between two ontologies from the two direct alignments of these ontologies to a reference ontology. The three anatomical ontologies under investigation are the Adult Mouse Anatomical Dictionary (MA), the NCI Thesaurus (NCI) and the Foundational Model of Anatomy (FMA). Methods: The direct alignment employs a combination of lexical and structural similarity. The indirect alignment simply derives mappings from direct alignments to the reference ontology. Each of the three ontologies is used, in turn, as the reference and evaluated against the other two ontologies. Results: Number of direct mappings identified: MA-NCI: 715, MA-FMA: 1,353 and NCI-FMA: 2,173. Number of indirect mappings identified through the reference: FMA: 703, NCI: 771 and MA: 741. Mappings specific to direct and indirect alignments are presented and discussed. Conclusions: This study confirms the feasibility of aligning two ontologies through a reference ontology. We also show that both the number of concepts and the number of concept names in the reference ontology are important parameters determining the suitability of an ontology to serve as a reference for deriving indirect mappings.
creating, integrating and maintaining multiple local ontologies is to adopt a global reference
ontology and a group of mapping rules between them. IF-MAP [
The Adult Mouse Anatomical Dictionary (MA) is a structured controlled vocabulary describing
the anatomical structure of the adult mouse [
The NCI Thesaurus (NCI) provides standard vocabularies for cancer research [
The Foundational Model of Anatomy (FMA) is an evolving ontology with an objective to
conceptualize the physical objects and spaces that constitute the human body [
1 http://obo.sourceforge.net/
We compare the direct alignment between two ontologies O1 and O2 to the indirect alignment
automatically generated from mapping both O1 and O2 to OR, the reference ontology. In practice,
we perform: 1) three direct alignments O1-O2, O1-OR and O2-OR; 2) the indirect alignment between
O1 and O2 through their direct alignments with OR; and 3) a comparison of the direct alignment
O1-O2 to the indirect alignment obtained through OR. In [
Our approach to aligning two ontologies directly first compares terms across ontologies lexically
in order to identify one-to-one concept matches. The second step is the identification of structural
matches. The interested reader is referred to [
The lexical alignment compares two ontologies at the term level, by exact match and after
normalization. Both preferred terms and synonyms in the two ontologies are used in the alignment.
For example, the concepts Heart valve in MA and Cardiac valve (synonym: Heart valve) in FMA
are identified as a match. Moreover, synonymy is used to identify additional matches. For
example, Cardiac valve in NCI and Heart valve in MA, although lexically different, are considered a
match because they name the same anatomical concept in the Unified Medical Language System®
(UMLS®) [
The structural alignment first acquires the inter-concept hierarchical relationships, IS-A and
PART-OF, and their inverses, INVERSE-ISA and HAS-PART, respectively. Missing relations are
generated through complementation, augmentation and inference techniques [
When a concept cR from OR is aligned with both a concept c1 from O1 ({O1: c1, OR: cR}) and a concept c2 from O2 ({O2: c2, OR: cR}), the concepts c1 and c2 are automatically aligned ({O1: c1, O2: c2}). The direct alignment MA-FMA identifies the match {MA: Heart valve, FMA: Cardiac valve (synonym: Heart valve)}, supported by positive evidence. The direct alignment NCI-FMA identifies {NCI: Cardiac valve, FMA: Cardiac valve}, also supported by positive evidence. Therefore, {MA: Heart valve, NCI: Cardiac valve} is derived automatically, through the FMA concept Cardiac valve, and supported by positive evidence in both direct alignments. 5
Results for three direct alignments are summarized in section A of Table 1. The alignment NCIFMA yielded the largest number of matches (2,173) and MA-NCI the smallest (715). A very small number of conflicts was identified in the two direct alignments to FMA; none in the direct MANCI alignment. In the three direct alignments, a vast majority of the matches (> 90%) was supported by positive structural evidence. No evidence (positive or negative) was found for 5-9% of the matches in three direct alignments. For example, although Elbow joint has relations to other matches in both MA (e.g., PART-OF Forelimb) and NCI (e.g., PART-OF Skeletal system), none of these paths are shared.
Results for the three indirect alignments are summarized in section B of Table 1. 703 matches between MA and NCI, 771 between MA and FMA, and 741 between NCI and FMA were automatically obtained from using FMA, NCI and MA as a reference, respectively. The majority of the three indirect alignments (88-92%) received positive evidence in both corresponding direct alignments they were derived from. 7-12% of them received no evidence and 0.4-1% received negative evidence in at least one of the direct alignments.
Taking the three ontologies pairwise, we compared the matches obtained in their direct alignment to the matches resulting from their indirect alignment through the reference. The results of these three comparisons are summarized in section C of Table 1. For MA-NCI, 654 matches are shared by both alignments, leaving 61 matches specific to the direct alignment (accounting for 8.5% of the direct matches) and 49 specific to the indirect alignment through the FMA. For MAFMA, 708 matches are shared by both alignments, leaving 645 matches specific to the direct alignment (accounting for 47.7 % of the direct matches) and 63 specific to the indirect alignment through the NCI. For NCI-FMA, 710 matches are shared by both alignments, leaving 1,463 matches specific to the direct alignment (accounting for 67.3% of the direct matches) and 31 specific to the indirect alignment through the MA.
88-89% of the shared matches in the three groups received positive structural evidence in all three direct alignments, e.g., {MA: Heart valve, FMA: Cardiac valve} in MA-FMA. Moreover, about 10-11% of the shared matches in the three groups received no evidence in at least one of the three direct alignments. For example, although linked to other matches in MA (e.g., HAS-PART Lung) and FMA (e.g., HAS-PART Ear), Body has no hierarchical paths to any other matches in NCI. This is why the matches of Body received no evidence in the two direct alignments MA-NCI and NCI-FMA, while receiving positive evidence in direct alignment MA-FMA. Lastly, nearly 1% of the shared matches in the three groups received negative evidence in one of the three direct alignments. For example, although a concept Nephron exists in the three ontologies, the corresponding match received negative evidence in the direct MA-FMA alignment (i.e., links to Renal tubule (synonym: Uriniferous tubule) through HAS-PART in MA but links to Uriniferous tubule through PART-OF in FMA), while receiving positive evidence in both direct alignments MA-NCI and NCIFMA. Domain knowledge is required to evaluate such matches. Alignment through a reference ontology is feasible and efficient. This study confirms the feasibility and efficiency of aligning two ontologies through a reference ontology. The proportion of matches from the direct alignment also identified in the indirect alignment is particularly good (91.5%) in the alignment MA-NCI with FMA as the reference. Assuming a good reference ontology is available, alignment through a reference is cost-effective: aligning n ontologies requires n(n-1)/2 pairwise mappings, but only n-1 mappings to a reference ontology. For five ontologies – which is a small number by Semantic Web standards – the difference already represents a 60% economy (4 vs. 10).
Suitability as a reference Ontology: Size matters. As shown in section D of Table 1, using
the FMA as a reference resulted in the identification of a vast majority (91.5%) of the direct
matches between MA and NCI. The large size of the FMA and its comprehensive set of synonyms
contributed to this high percentage of mappings [
Every ontology, large or small, contributes specific indirect matches. Regardless of its size, as shown in section C of Table 1, every ontology contributes specific indirect matches, i.e., matches that are not identified in the direct alignment. For example, using MA as a reference generated 31 specific matches, of which 19 received positive evidence in both direct alignments. For example, Glomerular capillary in NCI was not mapped directly to Glomerulus in FMA because the two terms are not synonyms in either ontology. However, the match {NCI: Glomerular capillary, FMA: Glomerulus} was identified indirectly when using the MA as a reference because Glomerulus and Glomerular capillaries are synonyms in MA. The match also received positive evidence in both direct alignments MA-NCI and MA-FMA. This indicates that the MA synonyms, although in relatively small number, play a significant role in the identification of mappings across two larger ontologies.
In summary, the most important finding of this study is that deriving an indirect alignment
through a reference ontology is not only feasible, but also reasonably efficient. Moreover, this
study confirms the intuition that both the number of concepts and the number of concept names in
the reference ontology are important parameters determining the suitability of an ontology to serve
as a reference for deriving indirect mappings. These findings are compatible with Burgun’s
“desiderata for domain reference ontologies in biomedicine”, including good lexical coverage, good
coverage in terms of relations and compatibility with standards [
Acknowledgements. This research was supported in part by the Intramural Research Program of the National Institutes of Health (NIH), National Library of Medicine (NLM) and by the Natural Science Foundation of China (No.60496324), the National Key Research and Development Program of China (Grant No. 2002CB312004), the Knowledge Innovation Program of the Chinese Academy of Sciences, and MADIS of the Chinese Academy of Sciences, and Key Laboratory of Multimedia and Intelligent Software at Beijing University of Technology. Thanks for their support to Cornelius Rosse, José Mejino and Todd Detwiler for the Foundational Model of Anatomy. Terry Hayamizu from the Jackson Laboratory contributed the evaluation of the direct mapping between NCI and MA.