RE: SemWeb BOF @ ISMB 04

> If I have understood you correctly, you are describing one of 
> the use cases for the Life Sciences URN identifier. A single 
> LSID can describe and give remote programatic access to both 
> your data (the gel data in the format you invented or some 
> other standard format appropriate to that data if one exists) 
> and  any amount of metadata in RDF that describes that data 
> and its cross-relationships with other entities. 

What I want is perhap something similar to a "context".  Let's use a very
simple hypothetical peptide sequence as an example.  If I have two peptides:
"MYLH" and "LEDA".  Each of them is a resource, so does each of the amino
acid that composed them.  Of course, if I want, I can assign each amino acid
a URI, but that will make the sequence data too verbose.  But if I don't
assign each AA a URI, how can I use RDF to describe the special property of
a particular amino acid within the peptide.  In other words, how can I
distinguish the Leucine of the first peptide from the one in second peptide.

Yes, theoretically we could invent an ontology to describe a protein
sequence so that each aimino acid could have a URI.  But then, to extend it,
why shouldn't we go further to the atomic level?  In that case, not many
people could even write a protein sequence without some professional help.
:-)

I roughly knows the idea of LSID (but will look in more detail).  But I am
not sure if LSID can sovle that problem.  But perhaps, one solution is to
register some sub protocol so that a URI can be implicitly assigned to
"part" of a resource. In a way similar to the MIME type where a program
knows how to deal with an application according to type, LSID might be able
to allow the implicit assignment of URI to fragment of a resource of a
particular type.  If this can be achieved, it will solve a lot of problem,
don't you agree? 

Xiaoshu

Received on Tuesday, 27 July 2004 09:18:35 UTC