[Ontology-editors] Precoordination of "response to" terms & drugs

[Chris Mungall]

If the referenced term in col 16 is a GO ID then there should be a  
separate co-annotation. See for example:
	http://wiki.geneontology.org/index.php/Annotation_Cross_Products#BP-MF_Example 
	

Whilst there is some redundancy, it ensures that tools can ignore  
col16 and be no worse off under the current system.

I'll try and make the docs clearer on this point

On Mar 13, 2009, at 8:07 AM, David Hill wrote:

> coannotation? we discussed this, but I don't remember what we decided.
>
>
>
> Gwinn Giglio, Michelle wrote:
>> That works, but then what about people coming to look for  
>> annotations to
>> "response to insulin".  Are we going to assume our users will  
>> properly use
>> and parse column 16 info?  I'm not opposed to making that  
>> assumption, but
>> depending on whether we assume that or not it changes how we decide  
>> to
>> structure things.
>>
>>
>> On 3/13/09 11:01 AM, "David Hill" <dph at informatics.jax.org> wrote:
>>
>>
>>> I thought we would annotate to response to drug and then put  
>>> insulin in
>>> column 16.
>>>
>>> David
>>>
>>> Gwinn Giglio, Michelle wrote:
>>>
>>>> Hi Chris,
>>>>
>>>> I  thought we decided that "response to drug" would not have any  
>>>> children.
>>>> People would annotate to "response to insulin" (for example)  
>>>> which would
>>>> have parentage under whatever proper chemical terms it usually  
>>>> does, then if
>>>> they new that in this particular case it was acting as a drug  
>>>> they could
>>>> co-annotate to "response to drug". I really don't think we should  
>>>> make
>>>> children of response to drug.
>>>>
>>>> If using column 16 I would expect the we would annotate "response  
>>>> to
>>>> insulin" and then we would need to somehow indicate that insulin  
>>>> in a
>>>> particular case was a drug - not sure how to do that.
>>>>
>>>> Michelle
>>>>
>>>>
>>>>
>>>>
>>>> On 3/12/09 4:55 PM, "Chris Mungall" <cjm at berkeleybop.org> wrote:
>>>>
>>>>
>>>>> What's our policy here?
>>>>>
>>>>> We already have a fair number of "response to X" terms.
>>>>>
>>>>> On the surface there is a potential for explosion here. You could
>>>>> imagine a high-throughput assay testing against a massive
>>>>> combinatorial chemistry library and measuring gene expression  
>>>>> levels,
>>>>> each experiment yielding an annotation where there is  
>>>>> upregulation or
>>>>> downregulation of genes. And if we go beyond chemical entities,
>>>>> there's a huge variety of behaviors an organism could potentially
>>>>> respond to.
>>>>>
>>>>> We can restrict the number of declared GO terms by applying our  
>>>>> rule:
>>>>> is response to X substantially different from response to Y? Are
>>>>> different receptors or pathways used? If not then one term will  
>>>>> do.
>>>>> The additional information can go in col 16. But it seems that  
>>>>> this
>>>>> may be hard for us to determine in many cases.
>>>>>
>>>>> Even so, I feel we should continue to pre-coordinate here. I don't
>>>>> fear the explosion here - this part of the graph can be managed  
>>>>> almost
>>>>> entirely automatically, like we are beginning to do with  
>>>>> regulation.
>>>>>
>>>>> This is related to the response to drug issue: some people don't  
>>>>> like
>>>>> the "response to drug" term, and we came up with a way of doing  
>>>>> this
>>>>> using annotation xps. But this maybe isn't necessary.
>>>>>
>>>>> CHEBI have moved from an overloaded is_a hierarchy to using  
>>>>> has_role
>>>>> relations between some entities and drugs. We can define  
>>>>> "response to
>>>>> drug" as "response to a chemical entity that is sometimes used  
>>>>> as a
>>>>> drug", and not worry about whether the gene product is acting in
>>>>> response to the chemical in its drug-role or non-drug role. is_a
>>>>> parentage under "response to drug" would be determined entirely
>>>>> automatically based on CHEBI.
>>>>>
>>>>> Personally I don't think "response to drug" is a great scientific
>>>>> term, but it seems it's useful for grouping and analysis  
>>>>> purposes, it
>>>>> doesn't really do anyone any harm. We could tag it in a slim.
>>>>>
>>>>> So I am thinking that precoordination is the way to go here.
>>>>>
>>>>> This arose from lakshmi's test GAF file with col 16 - they want to
>>>>> annotate to "response to ryanodine". Should we just suggest that  
>>>>> they
>>>>> request this term? Looking at the paper it wasn't clear to me
>>>>>
>>>>> Another question: should every "X receptor activity" term be  
>>>>> linked to
>>>>> "response to X (stimulus)" via part_of? I don't see why not, given
>>>>> current definitions.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>> _______________________________________________
>>>> Ontology-editors mailing list
>>>> Ontology-editors at geneontology.org
>>>> http://fafner.stanford.edu/mailman/listinfo/ontology-editors
>>>>
>>
>>
>
> -- 
> David P. Hill, Ph.D.
> Bioinformatics Scientist: Ontology Development
> Gene Ontology Consortium
> The Jackson Laboratory
> www.geneontology.org
> www.informatics.jax.org
> tel:207-288-6430
>
>