[MOBY-l] An Ontology question

[Benjamin Good]

Here is my brain dump for the day. Hope you have time to have a look and 
correct my misunderstandings :)

I guess maybe I'm a bit confused about the definition of "datatype" as 
it is used within the Moby world. If datatype refers simply to the 
structure and components of the objects, then why are there datatypes in 
the ontology that differ in their name but actually have exactly the 
same structure? For example from the ontology, phenotype_description isa 
object and hasa String (named Phenotype) and Interactor isa Object and 
hasa String (named role).

Clearly it is useful to be able to determine that a service operates on 
a "Phenotype" string and not a "role" string because it enables service 
discovery. For these particular objects, this semantic information is 
encoded only in the names of the objects. If the ontology was richer, 
perhaps the semantics could be spread across relationships between terms.

As I understand it, it is only really at the service discovery stage 
that semantics become important. With (hopefully) thousands of services 
out there, the appropriate selection of services is made possible by 
understanding the inputs and the outputs at a semantic level. The 
descriptions of these inputs and outputs could be stored in the objects 
themselves (as I have been envisioning), in the service ontology (as 
perhaps most others have already been thinking?), or (what do you think 
of this?) perhaps additionally through cross-references stored in the 
definitions of the services. For example, when a service operates on a 
specific class of data such as gene products “involved in seed 
development” that can be specified in an existing external ontology (for 
example by GO:0048316), the service definition could include a 
cross-reference to the pertinent node and authority. Such cross-refs 
would enable the use of external description systems for service 
discovery and facilitate the intermingling of moby clients with other 
applications.

?
-Ben





>>>>>>"Benjamin" == Benjamin Good <bmg at sfu.ca> writes:
>>>>>>            
>>>>>>
>
>  Benjamin> Hello Moby,
>
>  Benjamin> It seems that the moby system would benefit from a
>  Benjamin> semantically richer object ontology.  Has anyone thought
>  Benjamin> about ways to leverage existing ontologies (such as GO)
>  Benjamin> for this purpose?
>
>  Benjamin> curious to hear your thoughts 
>
>
>Well, GO itself describes the wrong thing, I think. MOBY needs an
>ontology of datatypes, and analyses. 
>  
>