Re: RDF for molecules, using InChI

Hi Egon,

A minor point:

You've created URIs such as http://pubchem.ncbi.nlm.nih.gov/#297 and  
http://pubchem.ncbi.nlm.nih.gov/#cid

Generally, it is considered better to mint URIs in a namespace you  
control or with which you have an agreement. Have you worked with  
Pubchem and do they sanction these URIs?

The downside of using URIs that are controlled by someone else is  
that you preclude the possibility of supplying meaningful behavior  
upon dereferencing these URIs. So according the the current  
recommendations, it is desirable that for a # uri, if one  
dereferences the URI before the #, with content negotiation set to  
ask for rdf, then one retrieves some rdf within which one would find  
further statements about the URI in question.

However if you ask:

wget -S --header='Accept: application/rdf+xml' http:// 
pubchem.ncbi.nlm.nih.gov/

You will see that you don't get rdf/xml. Changing that would require  
the cooperation of pubchem, hence my advised practice.

Similarly, the property iupac:inchi is an html page about inchis, but  
is used in your rdf as a property name. There are two issues here. 1)  
What sort of thing is iupac:inchi? We generally want to have a URI  
denote a single thing 2) Same issue as with pubchem. If you  
dereference http://www.iupac.org/inchi/ you won't get information  
about the property, like its domain and range.

There are reasonable alternatives to convey the same information -  
I'd be glad to discuss them if you are interested in doing so.
Certainly the links to blog posts about the compounds are a unique  
and valuable contribution!

Regards,

-Alan



On Aug 2, 2007, at 7:39 AM, Egon Willighagen wrote:

>
> Hi all,
>
> I played a bit with RDF for molecular data a bit this week, and now
> have a RDF provider service (try methane [1]), which is written in
> PHP, uses XSLT to create a HTML frontend (*). It works for any
> molecule/InChI, but depends on 'plugins' to set up any other than the
> implied properties (i.e. reproduce the InChI). The methane example
> mentioned shows some information extracted from Chemical blogspace
> [2], but I plan to write other plugins too, e.g. for PubChem,
> ChemSpider and other databases.
>
> I have written up some thoughs at [3], and would much like to hear
> your opinions and comments.
>
> Looking forward to hearing from you,
>
> kind regards,
>
> Egon Willighagen
> http://chem-bla-ics.blogspot.com/
>
> *) FireFox 2.0.0.6 and IE pick up the declared stylesheet, but
> Konqueror/Linux does not.
>
> 1.http://cb.openmolecules.net/rdf/?InChI=1/CH4/h1H4
> 2.http://cb.openmolecules.net/
> 3.http://chem-bla-ics.blogspot.com/2007/07/rdf-ing-molecular- 
> space.html
>
>

Received on Friday, 3 August 2007 08:46:34 UTC