Ontology Mapping Adaptation

The continuous evolution of life science ontologies requires the adaptation of their mappings to other ontologies. In this work, we proposed two approaches for tackling this problem in a largely automatic way: (1) a composition-based adaptation relying on the principle of mapping composition and (2) a diff-based adaptation algorithm individually handling change operations to update the mapping. Both techniques reuse unaffected correspondences, and adapt only the affected mapping part. We experimentally assessed and confirmed the effectiveness of our approaches for evolving mappings between large life science ontologies.

Groß, A.; Dos Reis, J.C.; Hartung, M.; Pruski, C.; Rahm, E. Semi-Automatic Adaptation of Mappings between Life Science Ontologies Proc. 9th Intl. Conference on Data Integration in the Life Sciences (DILS), Montreal, July 2013 2013-07

Datasets

Here we provide the datasets used for our evaluation. We evaluated the proposed approaches based on versions for three large life science ontologies: SNOMED-CT (SCT), NCI Thesaurus (NCI) and FMA. We extracted both, the source and the mapping versions, from UMLS. Thus, there can be differences between sources extracted from UMLS and original sources. Ontologies are provided in obo-format. The most important point in this work was to obtain the source versions as well as as their corresponding mapping versions, and to avoid preassigned differences due to inconsistent versions. For the experiments, we originally also extracted source versions from 2010 and 2011 that are also included in the dataset. Creating a more complete benchmark is an open challenge for future work.

We used the UMLS Metathesaurus to extract mappings between NCI and FMA (NCI-FMA) as well as SCT and NCI (SCT-NCI) from two UMLS versions 2009AA and 2012AA. We applied the mapping extraction method as described in this paper. For our paper, we adapted the mapping versions from 2009 with the proposed algorithms, and used the 2012 versions as reference mappings for evaluating the quality of the adapted mappings. It is important to notice that such reference mappings can be considered as a ’silver standard’, i.e., these mappings are not complete, and curators manually correct them by modifying also correspondences associated with concepts that did not underlie changes. For the evaluation of ontology mapping adaptation, we therefore eliminated such correspondences from the mappings since they do not change due to ontology modifications and can not be detected based on ontology changes.

The used ontologies and mappings are provided for download:

• Folder with all data files