[ debrief from Rachel, unscribed ]
<Joshua> caBIG & Semantic Web
scribenic: ericP
Joshua: i'm looking for collaboration ops with HCLS
<Rachel> thanks. Now I see everyone... :)
Joshua: caBIG semantic infrastructure:
... .. common data elements (for forms)
... .. also used inv info sharing structure
... .. UMLS models
... .. schemas for a SOAP-based info sharing infrastructure
... view-based integration (where views are ontologies)
... executed by query-rewriting
ericP: how much have you migrated data between RDF and e.g. SOAP messages
Joshua: we've annotated UMLS models linked to NCI thesaurus
... also wrote transformers to project data as OWL individuals
... going forward we need to provide access to linked data and SPARQL endpoints
[slide: Identifying Collaboration Opportunites]
<mscottm> tx
Joshua: next step is to apprach the folks in the CTMS space and look for use cases shared bwith HCLS
[slide: Ontology-aware Query Capabilities]
[slide: caBIG Use Cases]
Joshua: administration's focus on EMRs doesn't mean we should ignore e.g. integrated cancer use cases
[slide: CTMSi Scenario #9: Trial Matching]
Joshua: caEHR will be funded in the coming year
... build ontology-enabled EHR system
... e.g. decision support
Helen: everyone in your cancer centers can use your data model?
Joshua: yes, develop an EHR which can be deployed everywhere
... using BRiDGE and aligning HL7
... i.e. providing descriptions of the services
Helen: do you want to replace or bridge to existing systems?
Joshua: don't know the overall strategy, but if we can show how SemWeb handles DI, then SemWeb can help with this
Helen: yes, our view is to leave the data where it is and use a query transformation to get there
Joshua: with view-based integration, it's nice to have a shared view
... all caBIG work has been UML-oriented
... theoretically, everything in BRIDGE can be encoded in HL7 messages
... usefullness has not yet been demonstrated
... BRIDGE 3.0 is supposed to include an OWL representation
ericP: do you expect to need e.g. RDFS + transitive properties? or OWL individuals and disjoint classes?
Joshua: we need property mappings
... tools like d2r don't reflect the information artifacts in the information model
... it looks like you've overcome this with SPARQL CONSTRUCT expressivity
[slide: SW Advantages for CTMSi Scenario]
Joshua: for wider adoption in caBIG, we need to demo SW advantages
... .. literature suggests that expressing inclusion and exclusion criteria, we can e.g. see which are more inclusive, and guage efficacy
... Columbia research demonstrated bridging a semantic gap around brand names vs. generic names
<mscottm> slide9
Helen: when the mapping becomes complicated, we might hit infinite loops
... likewise on the drug ontology side, you can loop
[slide: SW Technology Maturity]
[slide: Potential Approach]
<mscottm> +1
[slide: ICRi Scenario #8: Personalized Medicine]
[slide: Scenario 8 - Overlay of Protein Array Data on Regulatory Pathways]
[slide: Potential Approach]
<mscottm> Helen: Do you have patient data?
<mscottm> Joshua: We could get it.