[BioRDF] from yesterday's discussion: "standard" vocabulary URI's

As a result of yesterday's discussion about "standard" vocabulary URIs, I
thought I would illustrate the point I was trying to make with an example.
The point I was trying to make concerned the difference between a URI that
references some RDF somewhere and a URI that references something that is
not RDF. In OWL, the former is accomplished with a object type property; the
latter is accomplished with a data type property. In my opinion, URIs for
any "standard/consensus" vocabulary terms should always reference some RDF
somewhere. Otherwise, I'm sure that some of the tools will break.

This example is based on a model that I have been carrying around in my head
about the details of one way to use RDF (in this case OWL) to integrate
databases (DB). Please note that:
1. this example hastily prepared to illustrate the point I was trying to
make yesterday
2. there are other models for integrating DBs using RDF
3. there are many variations on the model presented below
Also, since I am but a humble computer scientist, I do even begin to suggest
that I have made the correct disease name matchups in the example. Now, on
to the example.

Suppose you want to integrate OMIM and Brain Atlas DBs using MeSH disease
names. One approach would be to convert each of OMIM and Brain Atlas to
RDF/OWL knowledge bases (KB), and then connect these together with a bridge
KB that would link the OMIM/Brain Atlas disease names with the MeSH
vocabulary.

Here is the RDF/OWL for the example:

http://xlinux.sdct.nist.gov/jb/neuro-db-integration-model/vocab-kb.owl -
bridge KB

http://xlinux.sdct.nist.gov/jb/neuro-db-integration-model/omim-kb.owl - OMIN
KB

http://xlinux.sdct.nist.gov/jb/neuro-db-integration-model/brain-atlas-kb.owl
 - Brain Atlas KB

The bridge KB uses object type properties to reference the OMIM and Brain
Atlas disease names. There is also a data type property of type string that
references the HTML form of the MeSH entry for that name (e.g.,
http://www.nlm.nih.gov/cgi/mesh/2006/MB_cgi?mode=&term=Alzheimer+Disease&field=entry).

In the OMIN KB, the OMIM disease names are the index numbers into the OMIM
DB  with an rdfs:comment containing the title names of the OMIN DB entry. A
data type property of type string is used to contain the URL which will
return the entry's HTML form from the OMIM DB (e.g.,
http://www.ncbi.nlm.nih.gov/entrez/dispomim.cgi?id=168600). An object
property is used to reference the MeSH vocabulary name in the bridge KB.

In the Brain Atlas KB, the Brain Atlas disease name "Alzheimer_disease" is
part of several  links to alzheimer related images on the home page
(http://www.med.harvard.edu/AANLIB/home.html). Data type properties list two
of these links. Like the OMIM KB, there is an object property referencing
back to the MeSH vocabulary name in the bridge KB.

Although this exercise was not about the problem of relating different
vocabulary from different sources, it certainly illustrates the difficulty.

jb

Received on Tuesday, 23 May 2006 14:19:22 UTC