The challenges of developing an effective biomedical knowledge representation (KR) include (a) expressing the correct logic of domain-specific concepts in a reusable way, (b) excluding extraneous ontological commitments that complicate its use for a given task, (c) ensuring that the curation process scales, and (d) guaranteeing that it is understandable to domain experts. We are developing an ‘ontology design pattern’ (ODP, ) called the ‘Ontology of Experimental Variables and Values’ (OoEVV) as a pragmatic, reusable modular component based on earlier work , to capture definitions of variables and their values to be reused in other applications. We describe the design of the approach and tools we have constructed to support its use.
Gully A.P.C. Burns*, Jessica A. Turner
University of Southern California; ISI; Mind Research Network; UNM
Our formulation uses UML (Fig. 1). This includes Term, TermMapping, and Ontology classes to both import externally-defined ontology terms and to generate our own for our internal structures. OoEVV models consist of several elements: an OoevvvElementSet is a holder for all components relevant to a specific domain. This set contains ExperimentalVariable elements that each measure a defined quality Term (that may be drawn from any available external ontology). The mathematical properties of the variable’s values are specified by its MeasurementScale to provide a framework for managing the computations that may be performed on data measured with a particular variable. Scale subtypes include ‘Binary’, ‘Decimal’, ‘Integer’, ‘Nominal’, ‘Ordinal’ and ‘Hierarchical’ (to describe a hierarchical taxonomy). This UML-based structure is extensible to provide a practical methodology for developers to construct specialized data representations. OoEVV differentiates between ‘Qualities’, ‘Variables’ and ‘Measurement-Scales’ as its central core elements and provides a mechanism to capture different variables that measure the same thing. An example of this is ‘Handedness’ (PATO:0002201): a nominal scale might include three categories (‘left’, ‘right’ and ‘ambidextrous’, as with the children of the PATO term shown) whereas the ‘Edinburgh Handedness Inventory’ (obo:OBI_0001001) is a numerical scale providing a score derived from a questionnaire ranging from -100 to +100 (obo:OBI_0000991). We support best practices by providing good definitions and documentation, reuse of terminology wherever possible and compatibility with existing ontology formats and standards. We provide a practical curation toolset that may be used by domain experts to develop a structured lightweight terminology that may be accessed via the NCBO’s Bioportal.
We provide a command-line application (the ‘OoEVV’ toolkit, see http://www.isi.edu/projects/ooevv/overview) that uses spreadsheets to curate terminology and generate an OWL file that can then be uploaded to BioPortal.
 A. Gangemi and V. Presutti, (2009) “Ontology Design Pat-terns,” in Handbook of Ontologies, S. Staab and R. Studer, Eds.
 T. Russ, C. Ramakrishnan, E. Hovy, M. Bota, and G. Burns (2011), “Knowledge Engineering Tools for Reasoning with Sci-entific Observations and Interpretations: a Neural Connectivity Use Case,” BMC Bioinformatics, 12,:351.