[Go] 'binding issues' item listed on the GO Consortium meeting agenda

Harold Drabkin hjd at informatics.jax.org
Mon Mar 23 15:55:06 PDT 2009


comment stuck in below:

Chris Mungall wrote:
>
> On Mar 23, 2009, at 2:25 PM, Harold Drabkin wrote:
>
>> From one of Jim's earlier comments, I think a big problem is 
>> parentage: if Chebi has a relationship between, say glucose-6 
>> phosphate and glucose (child parent or something) , and we have a 
>> protein that binds glucose-6-phosphate, we don't want to imply any 
>> binding to glucose ( it might, but it might not).
>
> I think it's very important to mention the relationship type here. If 
> the relation is is_a, then I disagree
>
> If CHEBI says X is_a Y
>
> Then we are *forced* to accept that X binding is_a Y binding. If we 
> don't like this we can always request CHEBI change the relation.
Yes, this would be a problem;
>
> As I mentioned previously, with CHEBI's other relations all bets are 
> off. It is simplest to ignore these for now (in future we will want to 
> tackle how binding works in combination with part_of relations, but we 
> don't need to complicate the picture for now)
>
> This is extremely important. We should be working with CHEBI, not 
> making redundant conflicting statements.
>
>> Chebi may have relationships between, for example, for  ATP, ADP, and 
>> AMP, but a protein may bind very specifically to  one of them.
>
> If P binds specifically to X, and X is_a Y, then P binds specifically 
> to Y. this is built in to the is_a relation, there is absolutely no 
> getting around this.
>
> Fortunately CHEBI does not make an is_a relation directly between any 
> of these terms. Instead they are all is_a children of "adenosine 5' 
> phosphate"
>
> http://obofoundry.org/obo/CHEBI:15422+CHEBI:16761+CHEBI:16027.png
>
> (CHEBI calls this by the plural "adenosine 5' phosphate*s*", which is 
> wrong and will need to be changed in CHEBI. They are open to this)
>
> This means that if P binds to an ATP molecule, it is also the case 
> that P binds to an adenosine 5' phosphate molecule. Does this seem 
> reasonable?
 >> perhaps as long as it means "binds to some sort of adenosine 5' 
phosphate molecule", not any or all....


>
>> And I suppose there are many cases where a protein can bind related 
>> chemicals with different non-zero affinities, so that if one 
>> annotated binding to one there may be a good chance it would bind 
>> another. Should there be a rule in how one interprets the annotation 
>> line so as to not read any relationships among things in the target 
>> ontology to the actual single term designated as the target for the 
>> GO annotation? (hope this is clear).fcc
>
> I think that so long as the is_a relations in the external ontology 
> are correct, there is no danger of inferring false positives.
>
> This conservative approach means that we may have false negatives, but 
> that's always the case with GO.
>
>>
>> Harold
>>
>>
>>
>> Chris Mungall wrote:
>>> I have added this example to the col16 page:
>>> http://wiki.geneontology.org/index.php/Annotation_Cross_Products#Binding 
>>>
>>>
>>> I think the use of CHEBI in annotation-time cross-products poses 
>>> specific problems
>>>
>>> * It's often not clear (to me anyway) which CHEBI term to use. Jim's 
>>> example (linked to in the above page) has "PEP binding". CHEBI has 
>>> two terms in which PEP is a RELATED synonym: "phosphoenolpyruvic 
>>> acid" and "phosphoenolpyruvate". neither of these are defined in the 
>>> traditional OBO sense. It's not clear if the InChi strings count as 
>>> they are under RELATED synonym too.
>>>
>>> (Jim/Debbie - since you provided this example can you comment on 
>>> this, I'm too lazy to read the PMID, thanks!)
>>>
>>> * Many CHEBI terms do not have is_a parents. It's extremely 
>>> important for there to be correct is_a parentage. For example, if 
>>> the post-composed term is "phosphoenolpyruvic acid binding" then 
>>> this can be inferred to be a subtype of GO:0042301 ! phosphate 
>>> binding, based on is_a parentage in CHEBI.
>>>
>>> * CHEBI has other relations, but their semantics are unclear
>>>
>>> This gives me reason for caution. However, I am optimistic - CHEBI 
>>> are willing to fix these things in their ontology.
>>>
>>> In particular I would propose the following guidelines for use of 
>>> CHEBI in annotation time cross-products
>>>
>>> * The CHEBI term *should* preferably have a definition. The GO 
>>> annotator who uses a CHEBI ID in col 16 should propose a definition 
>>> on the CHEBI tracker
>>> * The CHEBI term *must* have is_a ancestry to CHEBI:24431 ! 
>>> molecular structure. If a GO annotator wishes to use a CHEBI ID in 
>>> col 16, and there is no such ancestry, the annotator should send a 
>>> request on the CHEBI tracker
>>> * The CHEBI term *should* have the correct synonym assignment. For 
>>> example, if the GO curator thinks the correct post-composed term is 
>>> "PEP binding" then PEP *must* be either an EXACT synonym or a 
>>> primary term name.
>>>
>>> On Mar 18, 2009, at 7:53 AM, Jim Hu wrote:
>>>
>>>> Hi Pascale,
>>>>
>>>> I think this is also potentially related to the discussion of 
>>>> whether/how the cross-product system can/should be used for 
>>>> post-composition of binding annotations.
>>>>
>>>> Jim
>>>>
>>>> On Mar 18, 2009, at 8:46 AM, Pascale Gaudet wrote:
>>>>
>>>>> Hi Emily,
>>>>>
>>>>> I am attaching the summary Debby Siegele put together about some 
>>>>> of the issues we've been having with substrate binding. (This was 
>>>>> sent to the ref genome list).
>>>>>
>>>>> Pascale
>>>>>
>>>>> Emily Dimmer wrote:
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> I noticed that the GO Consortium meeting agenda for Monday 30th 
>>>>>> March has a item entitled 'binding issues'. However there is no 
>>>>>> indication of who added this topic.Could someone please give a 
>>>>>> bit more detail about why this item has been added?
>>>>>>
>>>>>> I do agree that it is an area that is probably worth discussing 
>>>>>> (particularly with regards ISSing 'protein binding' annotations). 
>>>>>> Its just that I feel that agenda items which deal with annotation 
>>>>>> issues are more easily resolved when the meeting participants 
>>>>>> have had advanced notice as to the background of the discussion 
>>>>>> topic, so to be able to review their database's annotation 
>>>>>> sets/collect their thoughts!
>>>>>>
>>>>>> Thanks,
>>>>>> Emily
>>>>>>
>>>>>>
>>>>>>
>>>>>> ------------------------------------------------------------------
>>>>>>   Emily Dimmer Ph.D.
>>>>>>   GOA Coordinator
>>>>>>   EMBL-EBI
>>>>>>   Wellcome Trust Genome Campus
>>>>>>   Hinxton
>>>>>>   Cambridge CB10 1SD, U.K.
>>>>>>   Tel:     +44 1223 494654
>>>>>>   Fax:    +44 1223 494468
>>>>>>   email:  edimmer at ebi.ac.uk
>>>>>>   URL:    http://www.ebi.ac.uk/goa
>>>>>>
>>>>>> _______________________________________________
>>>>>> Go mailing list
>>>>>> Go at geneontology.org
>>>>>> http://fafner.stanford.edu/mailman/listinfo/go
>>>>>>
>>>>>>
>>>>> <BindingTermDocumentation.doc>_______________________________________________ 
>>>>>
>>>>> Go mailing list
>>>>> Go at geneontology.org
>>>>> http://fafner.stanford.edu/mailman/listinfo/go
>>>>
>>>> =====================================
>>>> Jim Hu
>>>> Associate Professor
>>>> Dept. of Biochemistry and Biophysics
>>>> 2128 TAMU
>>>> Texas A&M Univ.
>>>> College Station, TX 77843-2128
>>>> 979-862-4054
>>>>
>>>>
>>>> _______________________________________________
>>>> Go mailing list
>>>> Go at geneontology.org
>>>> http://fafner.stanford.edu/mailman/listinfo/go
>>>
>>> _______________________________________________
>>> Go mailing list
>>> Go at geneontology.org
>>> http://fafner.stanford.edu/mailman/listinfo/go
>>
>>
>



More information about the Go mailing list