Building scalable, query-able anatomy ontologies in OWL a question of balance
|David Osumi-Sutherland||University of Warwick|
Notice: Hosted by Vinay Chaudhri.
Date: 2012-08-06 at 14:00
Location: EJ228 (SRI E building) (Directions)
OWL has great potential as a language for representing anatomical knowledge and data in an intuitively query-able form. Biologists classify anatomical structures in many different ways including by structural attributes, developmental origin and function. Using OWL, we can make many of these attributes explicit as logical assertions and use the resulting assertions to automate classification and check for errors. These logical assertions can also serve as a substrate for queries about anatomy. Implementation of a standard abstract classification (upper ontology) can provide further constraints for basic consistency and sanity checking.
This approach is essential to scaling: maintaining an accurate, complete and non-redundant ontology with multiple axes of classification is impractical without auto-classification and error checking. But it is not sufficient. For an ontology to be useful to end users and maintainable by successive developers, logical formalization needs to be supplemented by extensive, human accessible textual content and careful attention to indexing so that relevant terms can easily be found.
Any formalization scheme also needs to be carefully balanced against other considerations: Is it sufficiently intuitive for ontology editors to work with? Will reasoning scale well enough as the ontology grows? Can each formal classification distinction be realistically made to fit an all/some pattern? Commitment to an upper ontology can provide useful sanity checking, but a focus on upper ontologies can lead to an obsession with distinctions that have no use to end users, or worse, run counter to end user requirements.
I will describe my attempts to achieve this balance in the context of modeling insect neuro-anatomy for the Virtual Fly Brain project.
David Osumi-Sutherland did his PhD at the University of Warwick on Xenopus development and a postdoc at Mass General Hospital on TGF-beta signaling in Drosophila development. On his return from the US to the UK, he pursued a career in science journalism, obtaining a Masters in Science Communication and publishing articles in a range of magazines, including Nature, while working part-time at the Drosophila genetics database FlyBase. Through his work for FlyBase he became interested in the challenges of knowledge representation in the context of communicating bulk bioinformatics data to communities of scientists. He is currently designing and building a model of Drosophila neuroanatomy for the Virtual Fly Brain project. He is also collaboratively developing a new upper ontology for anatomy and a general scheme for building arthropod anatomy ontologies.
Please arrive at least 10 minutes early as you will need to sign in by following instructions by the lobby phone at Building E. SRI is located at 333 Ravenswood Avenue in Menlo Park. Visitors may park in the parking lots off Fourth Street. Detailed directions to SRI, as well as maps, are available from the Visiting AIC web page. There are two entrances to SRI International located on Ravenswood Ave. Please check the Builing E entrance signage.