- From: William Bug <William.Bug@DrexelMed.edu>
- Date: Tue, 19 Sep 2006 17:30:34 -0400
- To: Matthias Samwald <samwald@gmx.at>
- Cc: systemsbiology hcls <public-semweb-lifesci@w3.org>
- Message-Id: <7C60E76F-0068-40D2-9866-48EFFFA9D2BD@DrexelMed.edu>
Hi All, Many thanks for providing you opinion on this important issue, Matthias. I would agree with Matthias on several points below and disagree on others. I would also add I'm am not the expert to debate this issue. The debate should be presented by those who've both curated/ developed these ontologies, as well as those who have experience using them to formally annotate instances of biomedical entities/ relations - formal representations that can be computed on effectively. This would likely include many who've posted to this list - including Matthias - as well as Alan Rector, Robert Stevens, Alan Ruttenberg, etc.. We'd certainly also want to hear both about the extensive work Barry Smith and his many colleagues have done on and with BFO, as well as the extensive work Nicola Guarino et al. have done on and with DOLCE. I would also add - as I mentioned in the call - there is a project out of INRIA en France that makes use of DOLCE to manage a neuroimaging repository and associated tools (actually contains physiological data as well such as EEG & EMG). That is NeuroBase: http://www.inria.fr/valorisation/rencontres/sante/transparents/ gibaud.pdf#search=%22neurobase%20dolce%22 There has also been extensive discussion on the issue of Upper Level ontologies at the Ontolog ontology community web site/Wiki: http://ontolog.cim3.net/ in particular see the Upper Ontology Summit held last Spring: http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit The OntoClean tool Alan R. mentioned was developed by Nicola Guarino, Chris Welty (http://ontolog.cim3.net/cgi-bin/wiki.pl?ChrisWelty), and their colleagues. As Alan Ruttenberg mentioned, it is very work reading the links on the following page, especially the application using specially descigned OWL classes to apply this methodology when using Protege-OWL: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntoClean One point I would stress here - though there are other candidates in this realm, most notably Cyc (http://www.cyc.com/cycdoc/upperont- diagram.html) and SUMO (http://ontology.teknowledge.com/) - it is BFO/ OBO-UBO and DOLCE (especially the application made of DOLCE components by Alan Rector, Robert Stevens, and Jeremy Rogers in their Simple Bio Upper Ontology) that are by far the front runner candidates for use in broad-scoped, field-wide semantically-aware biomedical data integration. The Cyc has been used very effectively by Peter Karp and his colleagues at SRI, but I don't know whether the foundational levels in Cyc really work for these large-scoped projects as well as these two other candidates (though I'm not the expert to debate this issue). In the end, as Matthias states, we are all - hopefully - describing the same biomedical reality, so one would hope to see all of these efforts converge. In practice, there will be many axioms and assertions across these foundational ontologies that simply will not be easily commensurable algorithmically (in some cases - not commensurable at all). The same will be true for anything one constructs on top of them. Please see below for additional comments. Cheers, Bill On Sep 19, 2006, at 4:07 PM, Matthias Samwald wrote: > > On the topic of foundational ontologies that are useful for the > biomedical domain, I feel urged to refer to the DOLCE foundational > ontology: > > http://www.loa-cnr.it/DOLCE.html > > DOLCE lite is available as a modularized OWL ontology. The OWL > version of DOLCE is already quite established and stable. true > The BFO SNAP/SPAN ontology and the relations ontology are almost > totally redundant with the basic modules of DOLCE. not true - at least not to my understanding of the relevant issues. As I mentioned in the call, there a basic assumptions about the what is being captured that are not commensurable either philosophically or mathematically from a "first principles" point-of-view. As far as the algorithmic ontology integration, I expect a great deal of code will need to be written to make both "worlds" commensurable - code that will be eminently brittle and error prone. I'd be glad to be proven wrong on this front by someone who's taken large biomedical data sets crossing many domains of knowledge (different modalities, levels of spatial resolution, and time resolution) and created a formal semantic representation for the content using one - then the other - foundational ontology, then been able to effectively integrate across them algorithmically. > The only innovations of BFO might be the two "transformed" and > "derives-from" properties, but these can also be added to DOLCE. > BFO also does not seem to offer any remedies to the shortcomings > that DOLCE is suffering from (especially the problems one runs into > when trying to express temporal indexing over properties like part- > of). not true - at least by my reading of the two, though I've only got a thorough working knowledge of BFO/OBR/OBO-UBO. There are fundamental differences in how the real world entities and relations are being formally encoded that are just not the same. Please don't take my word for it, though. Read: "BFO in a Nutshell: A Bi-categorial Axiomatization of BFO and Comparison with DOLCE" (http://www.ifomis.uni-saarland.de/Research/ IFOMISReports/IFOMIS%20Report%2006_2003.pdf) Hopefully, we can find a similar comparison performed by the DOLCE and/or GALEN folks. > As DOLCE is more extensive and has been very stable for some time, > while the OWL version of BFO is still under creation, I don't think > that there are any arguments for choosing BFO over DOLCE, at least > at the time. The VERY STRONG argument for using BFO/OBR/OBO-UBO is this will be the foundational ontology supported by the OBO Foundry. NCBO is explicitly committed not to taking sides on this issue, and I think that is wholly appropriate, given the various extremely important tasks they are charged with. The OBO Foundry group (Michael Ashburner, Suzi Lewis, Chris Mungall, Barry Smith, and others) - and the principles they are recommending MUST take sides on this issue - or to put it more accurately - must be able to host all OBO Foundry ontologies on top of a foundational layer this is algorithmically sound and supports a single view of biomedical reality. My hope would be this will ultimately accommodate all requirements including those DOLCE is particularly well suited to (e.g., decision support applications, etc.), but for now the foundation is BFO. > Fortunately, the redundandy between both ontologies should make it > easy to align them and settle this 'conflict' peacefully. Please see comments above. > > kind regards, > Matthias Samwald > > Bill Bug Senior Research Analyst/Ontological Engineer Laboratory for Bioimaging & Anatomical Informatics www.neuroterrain.org Department of Neurobiology & Anatomy Drexel University College of Medicine 2900 Queen Lane Philadelphia, PA 19129 215 991 8430 (ph) 610 457 0443 (mobile) 215 843 9367 (fax) Please Note: I now have a new email - William.Bug@DrexelMed.edu This email and any accompanying attachments are confidential. This information is intended solely for the use of the individual to whom it is addressed. Any review, disclosure, copying, distribution, or use of this email communication by others is strictly prohibited. If you are not the intended recipient please notify us immediately by returning this message to the sender and delete all copies. Thank you for your cooperation.
Received on Tuesday, 19 September 2006 21:30:49 UTC