[MOBY-dev] Re: OWL Lite encoding of bio-ontologies

[Anand Kumar]

Dr. Wilkinson, Dr. Suart,

I have received your email regarding OWL Lite encoding after being forwarded by
my colleague. I think the issues you have raised are very important in order to
have a full reasoner for OBO or in particular for GO. I can summarise the
problems at various stages.

1. Problem with OWL. OWL, being a DL based language, has all the advantages and
disadvantages of DL. While there are many advantges, one of the main weaknesses
of OWL over say, proper predicate logic is the weakness to represent
partitions, that is the aspects which can be used to classify classes or
universals. Also, there are issues regarding the fact that there exist
exceptions within biomedical relations. This topic has been discussed by Dr.
Alan Rector in this year's PSB. While I am not an open proponent of all the
solution he has (because of the inherent problems with DL and not because of
his good work), it could be useful to see his work at

http://www-smi.stanford.edu/projects/helix/psb04/rector.doc

2. Problems with the use of part-of within GO, and other OBOs. There is a very
good paper on this issue which shows where GO makes most of the "mistakes", and
that could be found at 

http://ontology.buffalo.edu/medo/isa.pdf

3. There are many other issues with GO, which are indirectly related to what is
the topic here and that has been dealt in

http://www.uni-leipzig.de/~akumar/Database_Integration.pdf

4. We are organizing a workshop at the University of Leipzig on these issues
specifically. Please attend, submit papers and encourage others to do so. The
details can be found at

http://www.ifomis.uni-leipzig.de/Events/GeneOntology_2004/

5. We are having a discussion, at the moment out of the lists regarding the
problems with multiple inheritance, where different aspects for classifications
are mixed together and the loss of data it brings. We could pursue that
discussion within the list or offline. Do you really think multiple inheritance
is a good thing? If yes, where specifically? If not, why not?

6. As for part-of. There is a lot of work especially within anatomy where
different part-of relations are mentioned. Foundational Model of Anatomy, one
of the most correct anatomy ontologies (though there are problems), has propoer
part-of, regional-part of, constituent part-of and systemic part-of. The paper
which describes them is at

http://sigpubs.biostr.washington.edu/archive/00000140/01/part_whole_amia2003.pdf

The idea of this email is not to provide a long to-read list. I have known your
work in the field of biomedical ontologies and would be particulary encouraged
if we could pursue this discussion and collaborate to find solutions to these
problems.

Kind Regards,
Anand.
-- 
Anand Kumar
IFOMIS
University of Leipzig
Härtelstraße 16-18
04107 Leipzig
Germany
http://www.uni-leipzig.de/~akumar/
http://labmedinfo.org/people/cv/kumar.htm

--
This message is from the GOFriends moderated mailing list.  A list of public
announcements and discussion of the Gene Ontology (GO) project.
Problems with the list?           E-mail: owner-gofriends at geneontology.org
Subscribing   send   "subscribe"   to   gofriends-request at geneontology.org
Unsubscribing send   "unsubscribe"  to  gofriends-request at geneontology.org
Web:          http://www.geneontology.org/