Saturday, July 25, 2009

ICBO Update

This week has been a busy one, with lots of events. I finished out the tutorials with a second day of events, powers, and dispositions, referent tracking, a WebProtege tutorial, and a discussion of outstanding issues with respect to Basic Formal Ontology, the Information Artifact Ontology, and the Infectious Disease Ontology.

The conference proper began yesterday. Highlights included a presentation by Stefan Schulz entitled "SNOMED CT's Ontological Commitment". It turns out that SNOMED CT is very unclear about its ontological commitment, which will hamper interoperability. Do so-called 'concepts' in SNOMED CT refer to entries in medical records, situations, or entities in reality?? One can find aspects of all three in SNOMED CT, and they're not clearly delineated and often mixed in the same hierarchy.

Another highlight yesterday was Werner Ceusters' presentation about the eyeGene research database. Again, the information model confuses entities in reality (the L1 Level of Reality), beliefs and hypotheses about L1 entities (the L2 Level of Reality), and information about L1 and L2 entities (the L3 Level of Reality). He recommends that the eyeGene database reform to make careful distinctions among them. And he believes it to be feasible.

A third highlight yesterday was Albert Goldfain's paper on the ontology of resistance, with special attention to methicillin-resistant Staphylococcus aureus (MRSa). He gave a detailed, precise, and high-fidelity representation of resistance.

Today's highlights included a presentation by Harold Solbrig, who gave an account of how he is trying to eliminate use of the word 'concept' in Biomedical Informatics and health care information technology. He nicely demonstrated how the word 'concept' leads otherwise reasonable people to cause all sorts of pathology in their representations (e.g., information models, data models, terminologies, etc). Although he has yet to convince people or the International Standards Organization (ISO) to eliminate the word altogether, by getting them to use 'ConceptDescription' in place of 'Concept', he has been able to limit the carnage.

Chris Mungall led off the day with an excellent paper "Cross-Product Extensions of the Gene Ontology". Because of combinatorial explosion, it is unwise to pre-coordinate every term that one might use from an ontology. However, by exploring the combinatorial space and isolating it from the core ontology, Mungall et al. were able to improve the quality of the core ontology as well as facilitate annotations with post-coordinated terms.

Melanie Courtot gave an excellent presentation (and got the biggest laugh of the day) on "MIREOT: The Minimum Information to Reference an External Ontology Term". Rather than duplicating terms from other ontologies, one wants to use terms from them when defining terms in one's own ontology. However, importing the entire ontology is impractical as they often are quite large. The MIREOT then, is the source ontology URI plus the source term URI plus the source term direct superclass URI. More software support is needed for this approach, but overall it is absolutely the right direction to go.

Later this evening, there is a poster session, dinner, and the plenary keynote. Tomorrow, more paper sessions and concluding remarks.

No comments:

Post a Comment