Re: Advancing translational research with the Semantic Web

On May 17, 2007, at 11:49 AM, Eric Jain wrote:
> How would you say e.g. "protein a is expressed in tissue b,  
> according to source c"?

For the demo, the way I said such things was like this (OWL available  
on request)

for cellular location:

There is a subclass of protein p each instance of which is located in  
some instance of tissue type b.
Then there was a property has_evidence who's value was an instance of  
the OBO evidence ontology.
That instance had a property called cites_evidence_source, which for  
the gene ontology  was typically an instance of journal_article that  
was identified by an instance of pubmed_record.

I need to write this up in more detail.

For the example you give, I might say something like:

There is a subclass of gene expression processes, during each  
instance of which some instance of protein a is the participant which  
is "the thing produced", and which is located_in some instance of  
tissue b.

I would then attach, as above, via an annotation property, the  
evidence instance that leads to the source.

The intended meaning is that the evidence says that  the class exists.

Note that this has problems (along the lines that Pat notes). For one  
thing, if you want to ask what the evidence was about, you need to do  
it in the asserted/told version of the ontology. That's because in  
the reasoned version of the ontology there may be additional axioms  
that become known about the class and one can't untangle them from  
the ones which the evidence suggested.

I don't think that other evidence being attached to the same class is  
problematic, though. However one needs to remember in this case that  
the class is not defined by the annotation properties which note the  
evidence - i.e. one evidence isn't about the other :)

I'm sure there are more problems, and I'm interested in figuring out  
how to resolve them.

-Alan