[MOBY-l] discussion of GO abbs's and Re: [MOBY] Constructing MOBY objects

[Mark Wilkinson]

On Tue, 2003-07-15 at 06:11, Beatrice Schildknecht wrote:

> So does each namespace registered in mobycentral first need to be 
> registered with GO? I was not aware of this, if it is indeed the case.

Well... yes and no.  We're in a transition state right now where the
MOBY project is trying to use the GO abbreviations with the aim of one
day offering to take over the burden of curating this list (and this is
in-line with other aspects of the MOBY project relating to our
cooperation with the I3C v.v. becoming a resolver for namespace
LSID's).  So... it isn't *necessary* for you to register namespaces with
GO, but it would be nice to keep the two lists synchronized until we
make the final step of becoming the "canonical" source for datatype
namespaces.


> This object will be a general phenotype description, that could traverse 
> all species. Ok, that makes sense--- a re-usable object. Is it possible 
> then, to alter/extend an already established object?

not only is it possible, it is a fundamental tenet of the MOBY project
:-)

> Will each Object namespace within the cross reference block have to be 
> registered separately? The number of namespaces will grow exponentially 
> if this is the case, as each and every piece of data could potentially 
> have an id of some kind.

yes, and yes.  I know, it sounds like we are heading for disaster!  But
it is fairly critical to do it this way, and hopefully as LSID resolvers
become more common the burden will become more dispersed.

> How are relationships between cross reference objects ascertained/Are 
> there relationships between cross reference objects? (Maybe this has all 
> been discussed elsewhere?? Is there a way to search the archives?)

You can search the archives if you click on the "mailing lists" link on
the main page and then go to the list you want.  

We have actually discussed this issue on the list before.  In
particular, a conversation between Heikko Schoof, Rebecca, and I several
months ago just before the 0.5 spec was solidified.  Heikko was hoping
to get cross-reference types into the spec, but I didn't feel
comfortable adding something so potentially important without taking the
time to think it through properly and coordinate our efforts with other
groups (e.g. GNF/Novartis) who are already making CV's of relationship
types.  I was already planning to raise this issue again soon so that we
could start thinking about it in anticipation of the next release...
Rebecca, perhaps you could tweak Heikko to come back to the land of MOBY
and re-state his opinions to get the conversation started?

> I am not, however, very clear about how the Invocation Class object 
> works, and how it is different from the Object within the CRIB. Looking 
> at the hypothetical examples in the API, I do not see differences in the 
> response.

The Invocation class is only necessary if your service changes the
namespace of the data as a result of the transformation.  e.g. if you
consume data in the NCBI_gi namespace and generate data in the
TAIR_Locus namespace then you need to put an Invocation object in the
Xref's that is in the originating namespace (NCBI_gi) to refer back to
the originating query object.  If you are doing, for example, a simple
lookup where you consume NCBI_gi's and produce Sequence objects in the
NCBI_gi namespace, then you don't need the Invocation object at all.

Cheers!

M