[Pankaj Jaiswal <pj37@co...>]

Using miscellaneous did come up but was negated in the group as it does 
not mean anything. We won't do it. To avoid such situations and those of 
creating the artificial terms like 'shoot part' etc. The best we think 
is terms that we certainly cannot classify in any known way is to put 
them as child of root term.

More later.

Pankaj

Jane Lomax wrote:

> Hi Pankaj - I would avoid having a term like 'miscellaneous' at all 
> costs, these sort of terms are bad ontology practice...how about you 
> just create terms like 'shoot part' as we've done in GO?
> 
> jane
> 
> On Tue, 9 May 2006, Pankaj Jaiswal wrote:
> 
>> I have just copied this mail to OBO-discuss and OBO-PONG list to keep 
>> an archive.
>> ------------------
>>
>> Hi Chris,
>>
>> When I started filling the missing is_a, I found about 25% of the 
>> terms didn't have the is_a as compared to the findings by OBOL. OBOL 
>> gave about 25-30 terms. Most of the problems with terms without is_a 
>> were solved by bringing in the "organ" term at the top in PO (it still 
>> needs populating though). However some terms as following have no 
>> known parent term defined in biology with which it can have is_a 
>> relationship. Reason being they are the only instance of that 
>> structure. Also we do not want to decompose them as we would like them 
>> to be as such.
>>
>> elongation zone (PO:0020125)
>> shoot (PO:0009006)
>> shoot apex (PO:0000037)
>> shoot node (PO:0005004)
>> phyllome (PO:0006001)
>> shoot internode (PO:0005005)
>> inflorescence (PO:0009049)
>>
>> So the options are to put them as is_a plant structure (root term), 
>> but the situation was not liked by the biologists as they do not like 
>> to see that many terms cluttering the top most level. They would like 
>> to see only those 7 terms on top as here
>> http://dev.gramene.org/db/ontology/search_term?id=PO:0009011. They 
>> claim that as long as part_of is there they are happy. The only query 
>> of a logical path they cannot complete is 'show me all the instances 
>> of a plant structure, that are is_a plant structure. Which is true 
>> based on relationship, but didn't we assign a aspect (subset: 
>> plant_structure) as well to say the same. I might be missing something 
>> here but somehow agree with the biologists.
>>
>> The other alternative was to create terms such as
>> -apical structures
>> -region/zones
>> -miscellaneous
>> and put the orphans under them with Is_a, but the concept of such 
>> terms is more artificial to them.
>>
>> Is there a solution.
>>
>> Pankaj
>>
>>
>>
>>
>>
>> chris mungall wrote:
>>
>>>
>>>> Hi Pankaj
>>>>
>>>> This is what people commonly do when importing GO and other OBO 
>>>> ontologies into other tools.
>>>>
>>>> One thing with this technique is that it's easy for errors to 
>>>> occur;  can you be sure that when you automatically create X is_a R, 
>>>> really  you should be creating X is_a Y is_a R (eg where Y is an 
>>>> existing  term). This would be the case if we were to do this with 
>>>> GO BP (eg  development terms).
>>>>
>>>> I would strongly encourage you to consider inverting the order of  
>>>> what you're doing, ie have the reference ontology be the correct  
>>>> (complete) one and generate the application ontology from that one.  
>>>> Or at least distribute the correct one on OBO etc. This isn't 
>>>> urgent,  but I think it's better to do it the correct way round 
>>>> first of all,  rather than switching later.
>>>>
>>>> Some POC specific reasons for doing this:
>>>>
>>>> you're going to want to take advantage of new DAG displays that 
>>>> allow  you to navigate according to relation - for anatomical 
>>>> ontologies  like PO with ontogenic relations it doesn't make much 
>>>> sense to  conflate all the relations together in the DAG display
>>>>
>>>> you're going to want to take advantage of the cross-product  
>>>> functionality of oboedit
>>>>
>>>> you're going to want to join the obo foundry
>>>>
>>>> here is a page showing what the de-convoluted DAGs look like. ( I 
>>>> can show you an example from POC too)
>>>> http://www.fruitfly.org/~cjm/obol/doc/dag
>>>>
>>>> There is a ppt there with some additional arguments for is_a  
>>>> completeness
>>>>
>>>>
>>>> On May 4, 2006, at 8:40 AM, Pankaj Jaiswal wrote:
>>>>
>>>>> Hi Jane and others,
>>>>>
>>>>> We had this discussion in the Plant Ontology's anatomy section on 
>>>>> filling up the is_a path.  Similar to your we were not successful  
>>>>> with a handful of terms because there was no way we could create a  
>>>>> parent for them. If we did want to create the parent we ended up  
>>>>> creating a non-biological or in other words a hypothetical 
>>>>> concept,  just to please the ontology gurus. The real biologists 
>>>>> were not  happy with this situation. So the decision was leave such 
>>>>> terms  without the is_a and use this file for actual browsing by  
>>>>> researchers for their work.
>>>>>
>>>>> However for ontology gurus and bioinformatics needs, we created  
>>>>> another file where a automated script creates a is_a parent for 
>>>>> all  the orphan terms. The parent assigned was the root term in 
>>>>> the  ontology itself. Thus giving them a file with complete is_a 
>>>>> paths.  BTW Lincoln came up with this idea and the group agreed.
>>>>>
>>>>> So with this reference, if we want  all the terms you have listed  
>>>>> to get is_a path assign the parent cellular_component (GO:0005575)  
>>>>> by default.
>>>>>
>>>>> It might look funny as there will be more number of terms in the  
>>>>> highest order but that's what will happen.
>>>>>
>>>>> Pankaj
>>>>>
>>>>>
>>>>> Jane Lomax wrote:
>>>>>
>>>>>> Hi everyone - attached is the new-look cellular component tree,  
>>>>>> with all is_a paths complete. part_of paths are also complete,  
>>>>>> with the exception of the annoying unlocalized protein complexes  
>>>>>> which we're working our way through (see https://sourceforge.net/ 
>>>>>> tracker/index.php? 
>>>>>> func=detail&aid=1475410&group_id=36855&atid=440764 and feel free  
>>>>>> to claim any! Or provide any advice on where they should go). I  
>>>>>> don't think we necessarily need to get these all done before we  
>>>>>> release
>>>>>> though, as currently none of the browsers have the capibility to  
>>>>>> show the is_a/part_of trees (although there is a prototype in OBO- 
>>>>>> Edit beta18, see below), but we certainly need to make fixing  
>>>>>> these a priority.
>>>>>> There may be a handful of redundant relations in there still -  
>>>>>> there's a slight
>>>>>> glitch with OBO-Edit's reasoner which I need to find them - it's  
>>>>>> fixed in the next release but I thought I may as well get this 
>>>>>> out  for comments before then.
>>>>>> So, comments please. If you want to view the alternate is_a/ 
>>>>>> part_of paths, launch OBO-Edit beta-18, turn on the reasoner  
>>>>>> plugin, restart OBO-Edit and the relation filter is the little  
>>>>>> icon at the bottom of the editing screen shaped like a little  
>>>>>> flower, second in from the right.
>>>>>> I'd like to announce when we're doing this in the first 
>>>>>> newsletter  - how much warning do you think we need to give 
>>>>>> people? A week?  Two weeks?
>>>>>> thanks,
>>>>>> Jane
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> -- 
>>>>> Pankaj Jaiswal
>>>>> G-15, Bradfield Hall
>>>>> Dept. of Plant Breeding and Genetics
>>>>> Cornell University
>>>>> Ithaca, NY-14853, USA
>>>>>
>>>>> Ph. +1-607-255-3103 / 4199
>>>>> fax: +1-607-255-6683
>>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>>
> 


-- 
Pankaj Jaiswal
G-15, Bradfield Hall
Dept. of Plant Breeding and Genetics
Cornell University
Ithaca, NY-14853, USA

Ph. +1-607-255-3103 / 4199
fax: +1-607-255-6683