[Biojava-l] ontology exception, addSequence & BioSQLSequenceDB

[Matthew Pocock]

>> -- Link a triple to a term, giving it a unique identifier. This allows
>> -- arbitrary expressions to be built up by then using the linked-to term
>> -- in another triple.
>> -- 
>> -- e.g. implies(isa(x, y), hasa(x, z))
>> CREATE TABLE term_relationship_term (
>>     term_relationship_id INTEGER NOT NULL,
>>     term_id INTEGER NOT NULL,
>>         PRIMARY KEY ( term_relationship_id, term_id ),
>>         UNIQUE ( term_relationship_id ),
>>         UNIQUE ( term_id ) );
>>
>
> I have 3 comments on this.
>
> 1) Given the unique foreign keys, this amounts to a 1-1 relationship 
> to both term and term_relationship, and therefore can be tacked on to 
> either of the two as a single foreign key. 

Yes. I didn't want to break the current schema, so added it this way.

>
> 2) It looks like you want to equate a triple to a term. Likewise, it 
> seems what you really want to have is a triple with subject and object 
> being triples themselves. 

Idealy, I want all triples to be terms themselves. Not all terms are 
triples though. The modified schema still supports 'anonymous' triples 
as well.

>
> 3) The example is poorly chosen and in fact I don't really understand 
> when you'd want all this. If x,y,z are arbitrary, then it can be 
> re-written as implies(isa,hasa), which is a 'normal' triple. 

This is different to your example because it's scoped by the type 'y'. 
Here is a more expressive rule - transitive closure (varables starting 
with '_' ):

implies(and(isa(_t, transitive), and(_t(_x, _y), _t(_y, _z))),
              _t(_x, _z))

Your example of imples(isa, hasa) is relying (I think) on some semantic 
magic - isa and hasa are both atoms, which happen to be relations. As 
such, they are not truth values that could be applied to the predicate 
implies. They only evaluate to truth values when applied to subject and 
object, but you're not specifying these - what do they default to?

Triples in expressions are not needed for pure data representation, but 
are needed (I think) for any encoding of rules (e.g. the hyperthetical 
rule: all EMBL feature types map to SO terms by looking up the SO term 
alias equal to the EMBL feature type).

Chris, am I barking up the wrong tree?

Matthew

>
>
> So I guess I'm missing something and I'd be glad to learn what.
>
> Also, Chris do you have any thoughts on this?
>
>     -hilmar
>