[MOBY-l] Objecs for ontology annotations

[Simon Twigger]

Hi there,

As a diversion from the RDF registration debate I have a question about 
expanding the existing object ontology to cope with different types of 
ontology annotations.

Our scenario is as follows: We have a variety of ontologies/controlled 
vocabs that we are using at RGD: GO, Mammalian Phenotype Ontology 
(we're working with MGD on this, see 
http://www.informatics.jax.org/searches/MP_form.shtml), a Disease 
Ontology (based around MeSH terms) and a controlled vocab for pathways. 
We're annotating genes/gene products and also Strains and QTLs with 
these terms.

These are all structured much like GO, the terms have an ID, a term and 
a definition and they are annotated to other objects with an evidence 
code and a reference, however, the evidence codes are different between 
the various ontologies.

We've been working on defining appropriate objects for MOBY services 
that handle these ontologies and related annotations to other objects 
and I wanted to see if others had addressed this same problem. Im not 
sure if Im thinking in an appropriately MOBY fashion and as a result 
getting too complicated.

Im still trying to write up what we have so it makes sense to an 
outside observer but I thought I'd ask to see if other people had ideas 
or if the Powers that Be (Mark?) could sketch out very briefly how they 
might solve this question?

Brief questions:
To connect an object and an annotation, Is it better to create an 
annotation object that contains the Annotated Object and the Annotated 
Data (eg an ontology term) and has attributes that define that 
annotation (eg. evidence code, reference, etc); or, should we subclass 
the annotated object and say it hasa annotation and add in attributes 
defining the annotation, eg. AnnotatedGene HASA GO term, and also HASA 
evidence code, reference, etc? At first glance the first option would 
seem more flexible than the second but I might be wrong.

We want to link these ontology terms to genes/gene products so I think 
we may need a 'Gene' object of some description. No-one will agree on 
what a gene is but would something lightweight like a GeneLocus object 
with just a symbol and name String attributes be acceptable?

Thanks for any ideas people may have!

Simon.

--

Simon N. Twigger, Ph.D.
Assistant Professor, Department of Physiology
Medical College of Wisconsin
8701 Watertown Plank Road,
Milwaukee, WI, USA
tel: 414-456-8802
fax: 414-456-6595
AIM/iChat: simontatmcw