[Ontology-editors] Disjointness [was Re: constituent part]

[Midori Harris]

On Thu, 7 Aug 2008, Chris Mungall wrote:

>
> On Aug 6, 2008, at 1:48 AM, Midori Harris wrote:
>
>> 
>> 
>> On Tue, 5 Aug 2008, Chris Mungall wrote:
>> 
>>> For now I'm maintaining the disjoint statements in a separate file and 
>>> running the reasoner in batch and generating reports.
>>> 
>>> Added a tracker item for:
>>> http://wiki.geneontology.org/index.php/Cellular_component_disjoint_classes#Extracellular_and_cell
>>> 
>>> I have not yet added a tracker item for:
>>> http://wiki.geneontology.org/index.php/Cellular_component_disjoint_classes#X_part_and_X
>>> 
>>> There is a section in the OE reasoner paper on disjoint_from (shared on 
>>> google docs under the geneontology account - let me know if you want 
>>> individual access)
>>> 
>>> Do you think the following example is a good one:
>>> ---
>>> In 2007 we made some high-level changes to the GO and introduced three 
>>> high level mutually disjoint classes within the biological process 
>>> ontology: cellular process, multi-cellular process and multi-organism 
>>> process.
>>> The editors used a configurable disjointness check in OboEdit to help 
>>> organize the graph such that these mutual exclusivity constraints held. At 
>>> first there were multiple violations of the disjointness constraint. For 
>>> example "conjugation with cellular fusion" was an is_a child of both 
>>> cellular process and multi-organism process. On closer examination the 
>>> definition of the class was found to be problematic, the definition was 
>>> changed and the disjointness violation was resolved by removing the is_a 
>>> link between 'conjugation with cellular fusion' and 'reproductive process 
>>> in single-celled organism'. [Can a GO editor check this? See email thread 
>>> Re: fixing non-disjoint violations Feb 5 2007]
>>> ---
>> 
>> This is an accurate representation of what we did, but biologically it's 
>> not very satisfying -- so much so that Val recently requested that the 
>> relationship be added back! We do need to improve the representation of 
>> sexual reproduction in single-celled organisms; as a stopgap, I've added a 
>> part_of relationship, which at least doesn't violate disjointness, but the 
>> area is still a bit ugly. It might not be a bad idea to pick a different 
>> example just for that reason.
>
> Any suggestions? Perhaps something that's come up recently in the process of 
> editing?

The only thing that's come up recently is revisiting 'conjugation with 
cellular fusion'; we've been pretty good about not violating the 
CP/MCOP/MOP disjointness constraints.

In the same email thread I found that Jane worked on 'cumulus cell 
differentiation':

   Basically, the graph looked like this:

   multicellular organism process
   ...
   ---[i] oogenesis stage
   ------[i] cumulus cell differentiation
   ------[i] fused antrum stage, oogenesis (sensu Mammalia)
   ---------[p] cumulus cell differentiation

   cellular process
   ---[i] cumulus cell differentiation

   so I just removed the is_a bewteen 'cumulus cell differentiation'
   and 'oogenesis stage' (not quite sure how it got to be both is_a and
   part_of oogenesis stage in the first place).

This is probably a better example because it wasn't at all confusing or 
controversial, and the outcome is biologically more satisfying than in the 
conjugation case.

m