- From: Christine Golbreich <cgolbrei@gmail.com>
- Date: Fri, 6 Mar 2009 10:01:29 +0100
- To: Ian Horrocks <ian.horrocks@comlab.ox.ac.uk>
- Cc: W3C OWL Working Group <public-owl-wg@w3.org>
2009/3/5 Ian Horrocks <ian.horrocks@comlab.ox.ac.uk>: > It may not be super useful/used in actual applications, but I don't see what > is wrong with our old friend the Narcissist for illustrating/motivating > local reflexivity. Clearly every instance of this class loves his/her-self, > but it is clearly not true to say that loves is reflexive in general. > > Ian > Thank you for your suggestion. As said, all the UCs selected in the document concern a) real needs and applications, and it was rather tricky to sort all the UCs to have b) for each feature 1 example from a UC easy to understand by users, c) with as many as possible domains involved e.g. chemistry inthis case. Such an 'invented' illustration of the feature, not grounded on a real needs and UC# , might reinforce other LC critics about having in OWL 2 new features which are not really useful or/and too costly. However if you wish I can add this within the text. IMO the auro-regulating-process is understandable by users of any domain, is as easy to understand as the "narcissist", thius does the jog Moreover, as you can see if you go to the document, at the end of that discussion with Michel, which aim was to get a real example from the chemistry domain, I had already implemented one of the chimical examples he proposed instead of the Ring structure (though IMO it was a more common notion like the regulating process, both from a specific domain but easily transferable to others...) Christine > On 3 Mar 2009, at 15:11, Christine Golbreich wrote: > >> 2009/3/3 Michel Dumontier <michel.dumontier@gmail.com>: >>> >>> >>> On Tue, Mar 3, 2009 at 5:05 AM, Christine Golbreich <cgolbrei@gmail.com> >>> wrote: >>>> >>>> 2009/3/3 Michel Dumontier <michel.dumontier@gmail.com>: >>>> - Show quoted text - >>>>> >>>>> >>>>> On Mon, Mar 2, 2009 at 3:24 PM, Christine Golbreich >>>>> <cgolbrei@gmail.com> >>>>> wrote: >>>>>> >>>>>> 2009/3/2 Michel Dumontier <michel.dumontier@gmail.com>: >>>>>>> >>>>>>> Hi Christine, >>>>>>> I understand that this is meant to be illustrative - its just not >>>>>>> particularly convincing as a use case, and it brings doubt to our >>>>>>> ability to >>>>>>> accurately model chemical knowledge. All atoms in a molecule are >>>>>>> (directly >>>>>>> or indirectly) connected to each other,irregardless of whether they >>>>>>> are >>>>>>> ring >>>>>>> atoms. Even if you wanted to say "SelfConnectedAtom" as an Atom that >>>>>>> isConnectedTo Self... what is the value in having such a class? There >>>>>>> is >>>>>>> none, in my opinion. >>>>>> >>>>>> First, you may have a property directConnectedTo (similar to >>>>>> directPart) and an axiom SubClassOf( RingAtom HasSelf( >>>>>> directConnectedTo)) that asserts local reflexivity for ring atoms. >>>>> >>>>> Yes, if you know that its a ring atom, you an certain do that, but >>>>> again >>>>> you're missing the point. being connected to itself is not particularly >>>>> interesting >>>>>> >>>>>> Do you mean that Cyclic Local reflexive isConnectedTo “Self” in Table >>>>>> 1 of your paper has no value either ? >>>>>> [1] >>>>>> >>>>>> >>>>>> http://sunsite.informatik.rwth-aachen.de/Publications/CEUR-WS/Vol-258/paper28.pdf >>>>> >>>>> As described in the text of the cited paper, we initially thought so, >>>>> but >>>>> without the rest of the solution (partial ordered paths over transitive >>>>> properties during reasoning), we don't get the intended result. so >>>>> that's >>>>> why we then tried rules, which works of course, but you have to specify >>>>> the >>>>> number of atoms in the ring you want to discover. >>>> >>>> OK, since that table should be considered as obsolete, I may use other >>>> chemical examples, which may be less contentious from a chemical point >>>> of view. However, >>> >>> Which table are you referring to? The one in the paper?? LOL.. you can't >>> conclude that from this discussion. >>> >>>> >>>> 1) As the UCs selected in the document concern real applications, it >>>> would be great if you might give a reference of the application where >>>> such class definitions are used, and even better a line showing the >>>> usefulness of local refmlexivity in reasoning (if you used it to >>>> reason) >>> >>> I gave you a use case, as you requested. Just because it hasn't been used >>> (yet) doesn't invalidate its prospective utility. >>> >>>> >>>> 2) >>>>> >>>>> eg A kinase that phosphorylates itself >>>>> Auto-Phosphorylating Kinase := >>>>> subclass ( Kinase hasSelf (phosphorylates) ) >>>> >>>> Not sure which syntax you use, but this axiom does not seem correct. >>>> To express what you mean, the pattern in FS should it not be instead >>>> something like: >>>> >>>> EquivalentClasses(Auto-p-A ObjectIntersectionOf(A hasSelf(p))) >>>> e.g. >>>> EquivalentClasses(Auto-Phosphorylating-Kinase ObjectIntersectionOf(A >>>> hasSelf(p))) >>>> Auto-Phosphorylating-Kinase are kinase that phosphorylates themselves >>> >>> sure, this could then be used to discover members of such a class. >>> there's >>> your example you asked about above. >>>> >>>> I may also simply replace present RingAtom example by: >>>> subclass ( Auto-Phosphorylating-Kinase hasSelf (phosphorylates) ) >>>> to express that Auto-Phosphorylating-Kinase phosphorylates themselves. >>> >>> fine >> >> OK, then not getting another real application UC from the group, I may >> use this to "illustrate" local ref, but we should be aware of its >> limitation i.e. >> 1) it has not yet been used in a real application >> 2) the same problems as those met with your RingAtom (global restrictions) >> hold. >> >> Christine >>> >>> -=Michel=- >>> >>>> >>>> best >>>> >>>> Christine >>>> - Show quoted text - >>>>>> >>>>>>> From the biochemical domain, proteins that sometimes modify >>>>>>> themselves - >>>>>>> some add phosphate groups in specific locations, and these proteins >>>>>>> are >>>>>>> therefore known as self-phosphorylating proteins. or certain RNA >>>>>>> molecules >>>>>>> will cleave themselves, and are known as "self-cleaving RNA" ... lots >>>>>>> of >>>>>>> other meaningful examples. >>>>>> >>>>>> Then if local reflexivity is useful, can you provide at least one real >>>>>> UC with an example in OWL2 which has value, to replace the ring >>>>>> example of UC#3 ? >>>>> >>>>> sure, >>>>> UC #XX - Capturing biochemical self-interaction as local reflexivity >>>>> overview: In Biochemistry, some biomolecules will chemical modify >>>>> themselves >>>>> in such a way that it has biologically important consequences. i) >>>>> Protein >>>>> kinases are enzymes capable of adding phosphate groups to certain amino >>>>> acids found within target proteins. Some kinases, known as >>>>> Auto-Phosphorylating Kinases, will add phosphate groups to certain >>>>> target >>>>> amino acids that are part of itself [1]. ii) Ribozymes are >>>>> catalytically >>>>> active RNA molecules in which 7 natural types are known to cleave their >>>>> own >>>>> RNA sequences. Such cleavage may result in significant changes to viral >>>>> replication, gene expression, and possibly the generation of different >>>>> protein transcripts. Such catalytically active, self-cleaving RNA make >>>>> up a >>>>> subclass of ribozymes called Self-Cleaving Ribozymes [2]. >>>>> >>>>> Features: Local Reflexivity >>>>> Example for: Local Reflexivity >>>>> eg A kinase that phosphorylates itself >>>>> Auto-Phosphorylating Kinase := >>>>> subclass ( Kinase hasSelf (phosphorylates) ) >>>>> eg A ribozyme that cleaves itself >>>>> Self-Cleaving Ribozyme := >>>>> subclass ( Ribiozyme hasSelf (cleaves) ) >>>>> references : >>>>> [1] http://www.springerlink.com/content/j36v22655088324r/ >>>>> [2] http://www.pnas.org/content/97/11/5784.full >>>>> >>>>> >>>>> >>>>>> >>>>>>> Now, that's not to say that the (bio)chemical work that i've >>>>>>> presented >>>>>>> doesn't have use cases for OWL2, its just that local reflexive has >>>>>>> just >>>>>>> not >>>>>>> yet been one of them. however, we have raised good examples of QCRs >>>>>>> (specifying the number and types of functional groups), reflexive >>>>>>> (hasimproperpart), asymmetric (hasproperpart), role chains (hasPart o >>>>>>> hasParticipant -> hasParticipant), disjoint union (all atom are one >>>>>>> of >>>>>>> the >>>>>>> atom types)... >>>>>>> one or more of these are much more interesting to present as use >>>>>>> cases >>>>>>> from >>>>>>> the chemical domain. i encourage you to consider these. >>>>>> >>>>>> For the other features we already have plenty of UCs and examples >>>>>> available, but I may keep UC#3 as yet another example of e.g. QCR >>>>> >>>>> great! >>>>> >>>>>> >>>>>> Christine >>>>> >>>>> >>>>> >>>>> -=Michel=- >>>>> -- >>>>> Michel Dumontier >>>>> Assistant Professor of Bioinformatics >>>>> http://dumontierlab.com >>>>> >>>> >>>> >>>> >>>> -- >>>> Christine >>> >>> >>> >>> -- >>> Michel Dumontier >>> Assistant Professor of Bioinformatics >>> http://dumontierlab.com >>> >> >> >> >> -- >> Christine >> > > -- Christine
Received on Friday, 6 March 2009 09:02:11 UTC