minutes for HCLS - Systems Biology

Here's what happened :)

Helena toured us through using Google Fusion Tables to generate a reaction
graph from a SPARQL query over a triple store containing the BioPax version
of SBML models in Biomodels database.

http://hcls.deri.org:8080/openrdf-workbench/repositories/SBML/summary

While that all works in principle, the problem is that the generated BioPax
does not contain URIs that can be integrated across models. More
problematically, is that it would be difficult to generate consistent URIs
from UnificationXrefs, because the "id" field contains some appended name:

<bp:UnificationXref rdf:about="xref_InterPro_IPR002394">
<bp:db rdf:datatype="xsd:string">InterPro</bp:db>
<bp:id rdf:datatype="xsd:string">BLL_IPR002394</bp:id>
</bp:UnificationXref>

I proposed that Biomodels generates BioPax using
identifiers.orgidentifiers for the UnificationXref and removes the
prefixed name to the
identifier. This would allow integration on those identifiers across all
the models.

I also suggested that we should look at Sarala's SBML2RDF work, which does
use identifiers.org URIs
https://sites.google.com/site/ricordowp4/file-cabinet/sbmltordf

We had a discussion about the goals of the work (to gain insight into the
network differences between physiologically different states), and how
biomodels would provide additional data beyond existing resources. So, we
resolved to compare the interactions in biomodels with those from iRefIndex
(available from Bio2RDF) and PathwayCommons.  I'm speaking with Gary Bader
today, so I'll bring up the question of whether they can generate biopax
models with identifiers.org URIs.  Scott also suggested WikiPathways as a
potential source of data.

m.

-- 
Michel Dumontier
Associate Professor of Bioinformatics, Carleton University
Chair, W3C Semantic Web for Health Care and the Life Sciences Interest Group
http://dumontierlab.com