[sbml-discuss] Testing Reactome export to SBML Level 2
Bernard de Bono
bdb at ebi.ac.uk
Mon Sep 6 13:22:21 PDT 2004
In Reactome, the same molecule may be represented in a number of different states such that each state has a distinct database ID. There is an inevitable loss in translating from Reactome to SBML data models, and I cannot rely on merely the name of the molecule to (a) reflect the unique state and thus (b) to have a one-to-one mapping with the database ID, particularly as I have to cull special characters from the name string during export to SBML.
Another important issue I wanted to cater for was the merging of two or more models from different sources - this required the Reactome export to establish a distinct ID namespace within the SBML model to facilitate this.
I worked under the assumption that the <List of Species> provided ample opportunity for modelling tools to set up a hash mapping from id to name. Clearly this is not the case - thank you for pointing this out to me. Is this behaviour shared by most modelling packages?
Dr Bernard de Bono
European Bioinformatics Institute
Wellcome Trust Genome Campus
Cambridge CB10 1SD,
----- Original Message -----
From: Robert D. Phair
To: 'SBML Discussion List'
Sent: Friday, September 03, 2004 8:31 PM
Subject: RE: [sbml-discuss] Testing Reactome export to SBML Level 2
I've tried out your SBML viewer and export functions on the Reactome development site. I downloaded several reactions and a pathway that I have some experience with and browsed the SBML. Everything appears correct.
I really like the collapsible-XML viewer.
Since we are also database people it doesn't surprise me that you have chosen to use the R_id in most places after the first definition of a species or compartment. This is good normal form. We've thought of doing the same thing with ProcessDB SBML output but have been troubled by the requirement it imposes on client tools that are not built on a database. For example, when I import your exported SBML into Herbert Sauro's JDesigner all reactants are labeled with their Reactome R_id only. Clearly the id is sufficient to *look up* the species name, but we've wondered if we should export the name instead, since the target app probably will be unaware of our unique ids. One could also argue that your exported SBML is sufficient on its own since it could be parsed to discover the name associated with the R_id, but we still wonder if it might be more useful to the SBML consumer if <speciesReference species = "R_113588" /> were rendered as <speciesReference species = "Fumarate" /> or perhaps the compromise adumbrated by your species name, <speciesReference species = "R_113588_Fumarate" />.
What are your thoughts?
Robert D Phair, PhD
Chief Scientific Officer
Integrative BioInformatics Inc
Rockville, MD 20854 USA
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Sbml-discuss