[sbml-annot] Annotation

[Neil S. <Nei...@ma...>]

Hi Team Annot,

Is anyone still here?!? And if so, did annotations ever get discussed at COMBINE in Paris?

I sent Brett and Pedro the following presentation to be gone through in my absence (I was on holiday, see slide 10) but I didn’t hear anything further.

https://www.dropbox.com/s/bls85iwe1vfwnw4/Annotation%20with%20RDF.ppt

My basic premise was that, while the previous Annotation proposal that we proposed a couple of years ago perhaps spiralled a little into the world of fantasy, the basic premise of updating the SBML annotation facility to take advantage of more features of RDF is a good one.

The usual response to claims such as this is to either a) suggest that I am a semantic web / RDF evangelist; and / or b) make some comment that RDF is too complex to support.

I personally don’t care less about RDF as such - I’m only interested in it in so far as that we’ve started to use it (since Core annotations were proposed) and to me it makes sense to continue rather than start again from scratch.

The second claim - that RDF is too complex - is as bogus as claiming that XML is too complex. Like XML, RDF is not meant to be written directly but accessed through a third-party reader/writer, like libxml, Xerces, or (in the case of RDF) Apache Jena.

The reason that I’m interested in this, is that there are a lot of things added to SBML (primarily as notes) which could be represented more formally as annotations with a minimum of extra work. There are examples aplenty in Recon2, path2models and others, but a simple example of annotating a metabolite with charge and inchi string is shown in the linked presentation.

To illustrate how easy it is to implement this with an existing RDF library (Apache Jena), a very simple example of a prototype is available here:

https://dl.dropboxusercontent.com/u/8980329/sbml-annotation.zip

This in Java / JSBML, but C++ RDF parsers are also available.

Finally, there’s also been a question about the overhead involved in supporting "all of RDF”. This certainly isn’t my intention. We can limit what we support by defining an RDF Vocabulary, which means that we control which predicates would be permissible in SBML annotations. It would work as a more formalised version of the current Biomodels qualifiers. An example of an RDF Vocabulary is Dublin Core, which we use already.

I’m pretty sure that this could be implemented fairly quickly if there’s need for this. Some features of this overlap with the proposals for the FBC package, but to me at least it would seem questionable to implement many different methods of annotating elements in different extension packages.

Cheers,

Neil.

Neil Swainston, PhD
Research Fellow

Manchester Institute of Biotechnology
University of Manchester
Manchester M1 7DN
United Kingdom