From: Chris M. <cj...@fr...> - 2005-11-04 20:02:37
|
According to the definitions in OBO_REL, "proteasome complex part_of cytosol" would mean that every particular proteasome complex would have to be part of some particular cytosol at some point in its existence. So I think GO is correct in not adding this relationship. On Fri, 4 Nov 2005, Jane Lomax wrote: > > actually - what about ubiquitous proteins that are found all over the > cell? I've just been having this discussion with David about > proteasomes...proteasomes are found in the ER, the cytosol and the nucleus > so we have proteasome complex as part of all these locations...I assumed > this was wrong because an individual proteasome can only be in one > location at one time...but what if that proteasome instance could move > between these locations? Presumably the would be relationship would be > okay? > > On Fri, 4 Nov 2005, Chris Mungall wrote: > > > > > Hmm, are they automatically wrong? We should take this over to the > > relations list (it gets a bit more subtle in when you factor in time). > > > > Whether or not they're automatically wrong, it may be a good way of > > spotting problematic part_ofs. I can produce a report of these (it'll be > > big!) and add it to the obo reports page, let you know when that's done. > > John, this might be an idea for an OBO-Edit check too (remember to take > > is-a into account here) > > > > I notice this is also a good way of noticing unusual cases where part_of > > crosses a boundary, for example "Golgi membrane part_of endomembrane > > system". Here we have part_of crossing a boundary from a component to a > > collection of components. > > > > On Fri, 4 Nov 2005, Jane Lomax wrote: > > > > > > > > Someone generated a list for us once (Joel from MGI??) that was all the > > > terms that had part_of relations to more than one parent - these are > > > automatically wrong with our use of part_of - perhaps we could generate a > > > another list? I know this doesn't catch all the incorrect uses of part_of, > > > but it's a good place to start... > > > > > > jane > > > > > > On Fri, 4 Nov 2005, Chris Mungall wrote: > > > > > > > > > > > Not automatically - this is down to curation and spotting the true path > > > > violations.. > > > > > > > > On Fri, 4 Nov 2005, Jane Lomax wrote: > > > > > > > > > Chris - is there a way you can figure out how many of these incorrect > > > > > part_ofs we still have and what they are? Might be good to make a > > > > > concerted effort to weed them out, > > > > > > > > > > thanks, > > > > > > > > > > Jane > > > > > > > > > > On Mon, 31 Oct 2005, Chris Mungall wrote: > > > > > > > > > > > > > > > > > [cc-ing to GO also] > > > > > > > > > > > > Hi Alan > > > > > > > > > > > > The GO usage documentation precedes the OBO relations ontology; see > > > > > > http://obo.sourceforge.net/relationship/ > > > > > > > > > > > > "type 2", necessarily part_of is essentially the same as the definition > > > > > > provided in the OBO relations ontology. The GO usage guide states that > > > > > > this the way that part_of is "usually" used. I believe that since this was > > > > > > written, most of the violations to the definition have been removed from > > > > > > the ontology. > > > > > > > > > > > > Note that the OBO 1.2 format allows you to further qualify part_of > > > > > > relationships between two classes, for example to state that A is part_of > > > > > > B and B has_part A where A and B are classes. Currently this feature is > > > > > > not used. This would be the same as placing two relationships, part_of and > > > > > > has_part between both A and B. > > > > > > > > > > > > At this time, neither of these strategies are being used in any OBO > > > > > > ontology. The main reason is that the part_of relation (as specified in > > > > > > the OBO relations ontology) holds far more commonly than has_part. > > > > > > > > > > > > In answer to your question: can we infer that for any A which is the > > > > > > object of the part_of relation for P1,P2,...,Pn, can we assume that A > > > > > > has_part B where B is one of P1,P2,...,Pn? I don't think so. Even if we > > > > > > were to assume that the knowledge provided in GO or any OBO ontology was > > > > > > complete (which is definitely a dangerous assumption), we still could not > > > > > > make this logical inference from the definitions provided. > > > > > > > > > > > > It may be still possible to logically infer that chromosome cannot have a > > > > > > has_part relation to the ER class from other relations, but this would > > > > > > probably rely on having additional spatial relations supplied in the > > > > > > cellular component ontology. > > > > > > > > > > > > Thanks for the interesting question > > > > > > > > > > > > Cheers > > > > > > Chris > > > > > > > > > > > > On Sat, 29 Oct 2005, Alan Ruttenberg wrote: > > > > > > > > > > > > > I've read http://www.geneontology.org/GO.usage.shtml#partof which > > > > > > > describes part_of and the relationship to has_part, and specifies > > > > > > > that in GO the meaning of part_of is that wherever the child exists, > > > > > > > it is as part of the parent. I understand that it is not necessary > > > > > > > that the parent has the child as a part whenever the parent exists. > > > > > > > > > > > > > > My question is about the possible values of has_part . Specifically I > > > > > > > am wondering whether one can infer that whenever the parent has a > > > > > > > part, that part must be one of those terms that is part_of the parent. > > > > > > > > > > > > > > Linking the question to the biological example in the documentation, > > > > > > > consider those terms which are part_of chromosome: > > > > > > > > > > > > > > chromosome, pericentric region > > > > > > > chromosome, telomeric region > > > > > > > chromatin > > > > > > > origin recognition complex > > > > > > > cohesin complex > > > > > > > provirus > > > > > > > Rad17 RFC-like complex > > > > > > > Ctf18 RFC-like complex > > > > > > > Elg1 RFC-like complex > > > > > > > Myb complex > > > > > > > > > > > > > > Would it be correct to infer that if a chromosome had parts, they > > > > > > > could only things on of the above list, and not, for instance, > > > > > > > GO_0005783:endoplasmic reticulum. If not, what lets us infer that an > > > > > > > endoplasmic reticulum can't be part of a chromosome? > > > > > > > > > > > > > > Thanks, > > > > > > > Alan > > > > > > > > > > > > > > > > > > > > > > > > > > > > ------------------------------------------------------- > > > > > > > This SF.Net email is sponsored by the JBoss Inc. > > > > > > > Get Certified Today * Register for a JBoss Training Course > > > > > > > Free Certification Exam for All Training Attendees Through End of 2005 > > > > > > > Visit http://www.jboss.com/services/certification for more information > > > > > > > _______________________________________________ > > > > > > > Obo-discuss mailing list > > > > > > > Obo...@li... > > > > > > > https://lists.sourceforge.net/lists/listinfo/obo-discuss > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > |