SV_MEETING_TITLE -- 30 Sep 2010

Link is for both the paper and the review

<ericP> http://www.w3.org/2010/09/30-hcls2-irc

<ericP> http://esw.w3.org/HCLSIG/PharmaOntology/Publications

Michel: Cover Matthias bigOWLim, then Eric/Chime, then paper comments

bigOWLim

<michel> Agenda (revised) TMO+LOD consistency checking - Matthias Patient data representation and reasoning - EricP and Chime Reconciliation of workshop paper with reviewer comments - paper & reviews - all next steps

Matthias: Uncovering inconsistency in ont, mappings or LODD data
... loaded data into TMO imports into bigOWLim repository
... reasoning rules, then queries, looking at subsumption issues
... happy to report that there are no inconsistencies so far!
... also need to find out about more soph. consistency checking rules
... looking into how to do more subtle checking
... default rules are called OWL max, mostly OWL Lite plus stuff, also a little RL
... Yes, loaded the mapping files from the wiki page
... but had to fix a few small inconsistencies there; should be an ongoing effort

Scott: Contacted OntoText?

Matthias: Yes, they would be glad to help w. consistency checking

Scott: There is a main OWLim programmer, ping Lucille
... Do other people on the list understand this agenda: importing into one repository, then run the reasoner

Matthias: Yes
... Wiki page from Matthias coming up...

<matthias_samwald> http://esw.w3.org/HCLSIG/LODD_TMO_consistency_checking

<mscottm> Scott tells Bob that Lucille is actually Vassil Momtchev :)

Michel: Does this include Silk mapping?

Matthias: Not sure about exactly which mappings

Michel: We have an LODD equivalence file that I will load in
... gives difference among provenance of source for drug data

TMO+LOD consistency checking - Matthias

Michel: Patient data representation

Patient data representation and reasoning - EricP and Chime

<ericP> https://dvcs.w3.org/hg/TMO-Indivo/rev/74371f75a6d3

Eric: Focus on mechanics, no fun this week
... recently committed change shows some files

<ericP> https://dvcs.w3.org/hg/TMO-Indivo/rev/74371f75a6d3#l2.2

Eric: tweaked xslt and started Indivo --> turtle xslt in link

<mscottm> SSL connection error.

Eric: 2 steps: reasonable coding in Indivo, then make sure turtle OK

<mscottm> me2

<matthias_samwald> me2

Eric: We have a problem w this link
... challenge is in Indivo, b/c xml extensibility problem, some of our tests are for SNPs
... have nothing to do w Indivo model
... trying to draw relationship between SNP and, say, BP test

Scott: SNP is like microarray data, also different from BP

Eric: Difficulty w XML extension for this stuff

Scott: Important to look at how SNPs are represented in RDF
... we should be partly prepared

Eric: Got it into RDF, hacked. Easy to do adequately, not too hard to do well

<ericP> http://esw.w3.org/HCLSIG/PharmaOntology/Queries#Q6._Are_there_any_AD_patients_without_the_APOE4_allele_as_these_would_be_good_candidates_for_the_clinical_trial_involving_Bapineuzumab.3F

Eric: it's the XML where the problem is, not RDF
... Look at this query 6

Scott: Should we try to harmonize w RDF for microarrays?

<ericP> <name>ADmark Alzheimer's Evaluation</name>

Eric: Depends on conventions for representing

Chime: Trying to extend XML to accommodate?

<ericP> Genotyping SNPs

Eric: Any medical record system taking these kinds of properties
... caBIG, i2b2, Indivo, another Harvard system are all medical systems

Anybody representing biomarkers in RDF?

Eric: We may be far ahead in attempting to do this

HL7, RIM, openEHR

<mscottm> Scott: Anybody representing biomarkers in RDF? There must be somebody.

<mscottm> Michel: Yes, we've done some work on it.

Eric: What is my mandate here? Started w Indivo.
... other piece is re-working the RDF side
... (discussing pipeline)

Scott: Let's not forget certain inform. needs will come up for provenance
... Ex, SNP from GWAS
... James Malone software ontology, may be useful here
... may be worth feeding back to Indivo folks

Eric: None of provenance in in Indivo originally
... but move provenance into RDF, hard-coded
... or leave provenance path for later?

Scott: Provenance could be by default according to which hospital or institute?

Eric: Still have to make something up here

Scott: As long as it creates the type of structure that we want
... extra bit of fantasy on RDF, might not be a real problem
... Can we ask the original people?

Eric: Chris gave us most of the medical perspective, probably had some institution in mind
... this would be the most realistic thing to represent

Scott: Clinician may by chance know the origin

More discussion of origins of data...

Scott: Will check out

Michel: XML --> RDF, etc, What more is there to do?

Eric: Looking for stuff appropriate for both us and Terminology

<matthias_samwald> (need to leave now, bye!)

<michel> bye!

<matthias_samwald> (by the way, i added an issue to the TMO issue tracker -- does anyone still look at that?)

<ericP> https://dvcs.w3.org/hg/

<ericP> https://dvcs.w3.org/hg/TMO-Indivo/

Eric: Everybody has access to this link
... but I have to push harder and harder to see updated changes...

<ericP> https://dvcs.w3.org/hg/TMO-Indivo/rev/74371f75a6d3#l2.2

Eric: here is stuff from trig generator, skeleton of what I'll add to

Michel: Q use of blank nodes(?) for patient; Needed to change this when loading

Eric: There was an 'empty url' problem instead: blank nodes should now be easier

Michel: Q is there inverse functional information here?

Eric: No. No national health code ID for linking patients
... they do a heuristic name search, etc. Don't need this quite yet?

Scott: This could be a real valuable feature!

Eric: But dangerous to put a stake in the ground here b/c HIPAA

Scott: Use something that could be used internally in clinic

Eric: We might endanger adoption of ontology if it relied on this

<mscottm> good

<michel> http://esw.w3.org/HCLSIG/PharmaOntology/Publications

Michel: Here is the link to the paper and the reviews

Michel: 1st review is happy, wants another figure
... figure would show patient --> drug --> experiment or test results
... make sure that upper level entities are represented, major types and relations

<michel> http://esw.w3.org/HCLSIG/PharmaOntology/Queries

Michel: Here is sparql query, want a graph to show the major classes
... how the data are interconnected, give a visual representation to sparql query
... review #2, wanted more detail about TMO function, put in one sentence about process
... but we did a good job about design process
... what is the coverage of the ontology? We did queries, and we answered all the questions!

Chime: Generally representative coverage

Michel: Took use cases etc, but would be nice to show a query that we actually could not answer
... Ex, doing regex queries
... highlighting this as an issue

Eric: TMO is intended as a connective tissue, not end-game
... but have had to move to clinical space
... that's the area where we will have insufficient coverage
... somebody who knows this space, need more feedback from these people

Michel: query #4 we already have a patient record aspect, using terms in other vocabularies

Eric: We may have pretty good coverage already
... clinical questions should cover standard stuff, plus test results

Michel: We did not show query to icd-9

Eric: snomed has some subsumption, we can show specializations with such queries
... does icd-9 have subsumption of any relations?

Michel: Tiimeline is good question!
... want to get all results by next week, so we can focus on the paper

<mscottm> Thanks Eric for the XSLT (i.e. gruelling) work!

<michel> Final Paper due November 16th

- DRAFT -

SV_MEETING_TITLE

30 Sep 2010

Attendees

Contents

bigOWLim

TMO+LOD consistency checking - Matthias

Patient data representation and reasoning - EricP and Chime

Summary of Action Items

Scribe.perl diagnostic output