Re: [CDAO-discuss] [Nexml-discuss] [Phenoscape] New page for discussion about top down inferences

[Jim B. <ba...@ne...>]

On Dec 9, 2009, at 5:50 PM, Hilmar Lapp wrote:

> I understand that there's no direct way to get those triples out, that's OK. But using an appropriate ontology (CDAO?), I would want to be able to use a reasoner that infers from a phenotype being associated with a state that the taxon that has the state is then also, by some relationship, related with the phenotype.
> 
> I understand that that's not what the XSLT is doing - rather it hard-codes that logic I suppose. But if we can use a reasoner and an ontology that declares the semantics of matrix, state, TU, and character, we would supposedly not need to hard-code that but simply run the reasoner to infer the taxon-phenotype relationship from the state-phenotype relationship.
> 
> Am I off-track here?

I think you could provide a relation chain via the ontology you're using which would let the reasoner make that direct link.


>> We need a way to embed what the actual link between the taxon and the state/phenotype is (because now we have multiple options).
> 
> I see. So the relationship between state and phenotype does now no longer unambiguously determine the inferable relationship between taxon and phenotype. Is that what you are saying?

Yes, we currently provide a specific link between the phenotype and the state.  But I think that link is the same regardless of which link we want to exist between the taxon and the phenotype.  Also, for a given state and phenotype, we want to link different taxa to that using different relations.  (For everyone who may have no idea of the context - we want to distinguish different intents when a phenotype is associated with a higher level taxon and meant to apply to all subtaxa, vs. another weaker association that doesn't down-propagate in that way).

- Jim