[Gusdev-gusdev] Re: proposed modifications to GO Schema for GUS3.0

[Sharon D. <di...@SN...>]

Hi Arnaud,

Thanks for the quick review.  

Arnaud Kerhornou wrote:

> Hi Sharon
>
> I had a discussion with Matt and from our point of view, the design 
> you're proposing sounds sensible.
>
> Just two clarifications about evidences:
>
>    (1) for an evidence attached to a GOAssociationInstance, it can be:
>        *  Similarity data,
>        * a AAMotifGORule,
>        * a DB Reference, stored in an ExternalDBEntry if it's external,
>        * if it's a pubmed_id, we may want to store the whole reference 
> into a reference entry. But I guess it can be done with the schema 
> you're proposing, is that right ?


That is correct.  Since we're using the generic Evidence table, the 
"fact_table" could be any table in GUS.  There are no constraints that 
force you to have all the DB refereneces stored in an ExternalDBEntry.

>
>    (2) for an evidence attached to a GOAssocInstEvidenceCode as it may 
> be an "internal DB link", e.g. a protein_id for example, the DB name 
> prefix will be GeneDB in our case, or if it's a GO_id, it will be GO 
> as a prefix. Will it be generated on the fly ?


I would view this as something generated on the fly.  I suppose for 
external databases (such as GO), we could add an attribute to the 
ExternalDatabase table that represents the "abbreviated name".  This 
abbreviated name could then be used to prefix identifiers.

>
> Also re. the GOTerm table, would it be possible to add a comment field ?

I do not have a problem with this.  Does anyone else have an opinion on 
this?  What type of information would you put in the comment field?

Cheers,
Sharon

p.s.  I will be out of town until Monday July 22nd.



>
> cheers
> Arnaud
>
> Sharon Diskin wrote:
>
>> Hi Arnaud,
>>
>> I've attached the new GO schema and documentation that Jonathan Schug 
>> and I have come up. We believe that this new schema addresses your 
>> concern about the association date and also provides a better way of 
>> tracking multiple instances/sources of the same association (see 
>> description of GOAssociationInstance table).
>>
>>
>> Regarding your other question concerning the 'DB:Ref' and 'from or 
>> with', we see these being tracked using the generic Evidence table.   
>> The 'from or with' would be Evidence for the GOAssocInstEvidCode 
>> entry as it is connected to a specific evidence code that is assigned 
>> to the association.  The 'DB:Ref' (such as in the PubMed example you 
>> mentioned) would be Evidence for the GOAssociationInstance entry as 
>> it is _not_ connected to a specific evidence code, but rather the 
>> association as a whole.  More details can be found in the attached 
>> documentation.
>>
>>
>> Any feedback is welcome.  Also, if any of this is unclear, just let 
>> us know.
>>
>> Cheers,
>> Sharon
>>
>>
>
>
>