HCLS Teleconference

23 Jul 2009

See also: IRC log


AlexPassant IPcaller JimS Julia Julia_Kozlovsky KevinDoyle MikeHucka Oliver SusieS curoli ericP



ericP: we've been remiss on task updates -- aim to fix that in the near future.
... please remember to send mail to public-semweb-lifesci instead of long lists of individuals.
... helps others keep track of what's going on and potentially provide input.
... helps us motivate others to work with us.
... can prefix with [BioRDF] or whatever so folks can filter (on or off) if they want.

SBML & the Semantic Web - Mike Hucka

mhucka: SBML focuses on modelable/simulatable processes
... before 2000 there was no common model for exchange between software
... no limitation on the expressivity
... not procedural, so doesn't tell you e.g. how long to run
... some people encode abstract models which are (intentionally) not bilogically feasible
... the MathML does not support integrals
... right hand side is a diff eq and left hand side is the rate of change of the species
... events model things like a cell reaching a certain size and undergoing mitosis
... "validation" not all captured in XSD or UML, includes some SBML-specific pragmas
... evolved some process, e.g. editors elected for 3 years
... still adding formalism to SBML
... level2 added constructs like initial assignment
... level3 components for things like shape
... SBML vars correspond to e.g. vars in papers
... Miriam provides (amoung other things) a minimal provenance model
... Miriam is not SBML-specific, but SBML provides a Miriam binding to map Miriam concepts to SBML's expressivity.

Oliver: do you have to select the resources from the miriam list?

mhucka: for longevity, you want to have confidence in the URL
... had some troubles with LSIDs
... added a redirection
... e.g. GO has changed over the years

Oliver: would i use a Uniprot ID or a Miriam ID?

mhucka: encourage urn:miriam:... to abstract location
... bqbiol = AKA

JimS: we have RDF beyond basic Miram -- e.g. tying to BioPAX
... can we add what we want?

mhucka: Miriam-SBML is minimal
... at dev time, many langs didn't have RDF parsers
... you can add whatever, but there's no guarantee that tools will understand it
... put your additional RDF after the SBML-controlled bits

[BioModels advanced search page]

[slide 31 is a screenshot of the BioModels advanced search page]

ISWC Workshop Update - Susie

susie: wrote a blog and email nagging folks to submit papers
... deadline in 10 days

SBML & the Semantic Web - Mike Hucka

[slide 32: bleeding edge]

mhucka: should also include sybil (SP?) - integrating BioPAX and SBML
... first bullet: rich anntations enhance identy and enable model merging
... second bullet: MFO accesses Systems Biology Ontology
... third bullet: paper at last ISMB

[slide 35: kudos and upcoming meeting]

Oliver: re: extra RDF, is the parser generic RDF?

mhucka: some folks would like to change that
... may have been a design flaw, but goal was to make it easy for tools

ericP: i think it was a good engineering soln. will want to lift constraint when folks want to use generic serializer

<mhucka> SBML Level 2 Version 4 specification document: http://precedings.nature.com/documents/2715/version/1

<mhucka> Section 6 describes the RDF bits, see especially p. 93

ericP: do you have RDF-like extensibility in the rest of SBML?

mhucka: extensibility point is in the annotations
... the SBML Level 3 packages that layer on the core, add elements and attributes to add representational power

<KevinDoyle> I have to split. Thanks for all of the info.

ericP: if you use these modules, should a naieve tool still interpret core?

mhucka: there's a mustUnderstand flag in the packages
... mayUnderstand e.g. auxilliary picture
... a tool interpreting the multi-package needs the package for complete understanding
... a level3 hierarchical model could be flattened into level2

ericP: interop seems like it could happen two ways:
... .. SBML->RDF via GRDDL
... .. RDF->SBML via SPARQL Query + XSLT
... this would be motivated by us wanting to take advantage of your data or you wanting to use our data

mhucka: what sort of data?

ericP: pathways, receptors, GO
... do you include the relationships between GO terms?

mhucka: not the relationships
... an integrating query would have to make sense to some consumer
... many SBML users are interesting in time-series simulation
... maybe model creators who want to draw from other data sources
... pathway expression use case would be interesting -- happening in Sybil (Oliver)
... semantics-heads are trying to merge and verify models (U Newcastle)
... the models are getting more complex and prohibitive to construct from scratch
... composition will get more useful as models grow

ericP: can you envision SMBL being completely in RDF? [brief MathML in RDF discussion] a motivating drug discovery scenario would be a researcher leveraging SBML pathway models when testing if drug X could do Y

mhucka: prolly could be all RDF. could then model more of the constraints
... Allyson Lister MFO was doing that

Oliver: we're mirroring the core of SBML to RDF
... project: SBPAX

ericP: noting that you provide urn resoultion, did you consider PURLs?

mhucka: some detail made it easier to create our own resolution service
... long questions about what would be good to persue

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2009/07/24 18:17:11 $