Re: [HCLSIG] homologs for yeast proteins?

chris mungall wrote:
> I disagree - annotation using particular terms are exactly the kind of 
> thing you need semantic web magic for. In this particular case 31930 has 
> no is_a or part_of children, so extra magic would buy you nothing; but 
> for other queries you absolutely need to do inference to get other 
> implied terms; and this inference will have to be more sophisticated as 
> more relations from the OBO relations ontology are used in other OBO 
> ontologies, and as more annotations are made by composing anonymous 
> classes from a mixture of OBO ontologies.

You can implement a system that supports queries on is_a and part_of 
relationships or that maps from one data set to another without any 
semantic web technologies. For example, go to http://www.godatabase.org/ 
and browse by term. I'm quite sure they don't have a triple store with an 
inference engine backing that site :-)

The main argument in favor of using more sophisticated technology here 
isn't that you couldn't achieve the same functionality by other means, but 
that you can achieve the same functionality with less effort.

Now in this particular case I'm not sure you'd manage to convince them that 
the benefits outweigh the effort required to adopt this technology, 
especially not since the structure of their data isn't that complex anyway.

But by providing their data in a semantic web compatible way, they make it 
easier for (and therefore encourage) other people to integrate their data 
into their own applications (if they are enlightened enough, that is ;-).

Received on Tuesday, 31 January 2006 23:11:48 UTC