HCLSIG/PharmaOntology/Meetings/2010-05-13 Conference Call

Conference Details

* Date of Call: Thursday May 13 2010 
* Time of Call: 12:00pm - 1:00pm ET 
* Dial-In #: +1.617.761.6200 (Cambridge, MA) 
* Dial-In #: +33.4.89.06.34.99 (Nice, France) 
* Dial-In #: +44.117.370.6152 (Bristol, UK) 
* Participant Access Code: 42572 ("HCLS2") 
* IRC Channel: irc.w3.org port 6665 channel #HCLS2 (see W3C IRC page for details, or see Web IRC) 
* Mibbit quick start: Click on mibbit for instant IRC access
* Duration: 1h 
* Convener: Susie

Agenda

• TMO Updates - Michel, Elgar
• Notes Functionality - Trish
• Ontologies of Interest (OCR, OMOP, OGMS) - Trish, Joanne
• Indivo Schema - Michel, EricP
• IO Informatics Data - Elgar
• Sources of Patient Data - Scott
• Provenance Requirements - Joanne
• Interface/eMerge - Bosse, Chris
• Outreach (IHI, Bio-IT World Europe, UPenn Translational Medicine) - Michel, Susie
• AOB

Minutes

Attendees: Chris, Janos, Julia, Joanne, Bob, Scott, Elgar, Trish, Jun, Bosse, Christi, Michel, Paul, Susie

* <Susie> TMO Updates
* <Susie> Elgar: Starting digging into drug ontologies and RxNorm
* <Susie> Elgar: Have been working on that part of the ontology
* <Susie> Elgar: Will have more in 2 weeks
* <Susie> Elgar: Will be in touch with Janos
* <Susie> Trish: TMO is in BioPortal
* <Susie> Trish: The google code svm will be the main repository
* <Susie> Trish: There isn't an automatic update from google's svm
* <Susie> Trish: Will check with Elgar and Michel to get that squared away
* <Susie> Scott: TMO is available from a SPARQL endpoint
* <mscottm> NCBO ontologies are available from this experimental endpoint (since yesterday): http://sparql.bioontology.org:8080/sparql/
* <mscottm> http://dev.adaptivedisclosure.org/search/?server_url=http://sparql.bioontology.org:8080/sparql/&repository_name=ht...
* <Susie> Scott: Should be able to access the ontology directly
* <Susie> Scott: from the link that I just pasted
* <mib_0gqank> Trish: NCBO will be adding documentation 
* <Susie> Scott: It's different in nature from the BioPortal interface
* <Susie> Trish: The triple store back end is still experimental
* <Susie> Trish: Do want to collect feature requests
* <Susie> Trish: But they will be in a holding pattern for a little bit
* <mib_0gqank> and a support email to collect requests, however this is a prototype and requests will not be immediately addressed
* <mib_0gqank> http://bioportal.bioontology.org/visualize/42758/?conceptid=transmed%3ATMO_0016
* <Susie> Notes Functionality
* <Susie> Trish: Notes allows comments about particular bits of the ontology
* <Susie> Trish: NCBO and Protege have been working to combine an infrastructure for that
* <Susie> Trish: So Protege, and BioPortal use the same framework
* <Susie> Trish: They point to the same backend
* <Susie> Trish: So new term can be proposed in BioPortal 
* <Susie> Trish: And can then be seen by the ontology experts in Protege and added in the right place
* <Susie> Trish: Structure is being added to the notes
* <robertpowers_> Susie: Notes sounds really useful
* <Susie> Trish: Notes available through RSS feeds and web services
* <Susie> Trish: In the future, if you’re in an annotation application and want to submit a note it'll be possible to do through the web service
* <mib_0gqank> http://www.bioontology.org/wiki/index.php/Ontology_Notes
* <Susie> Trish: Would like to hear from people who are interested in using that functionality
* <Susie> Trish: Helps to bridge the 2 communities
* <Susie> Other Ontologies
* <Susie> OMOP
* <Joanne> OMOP Observational Medical Outcomes Partnership
* <Susie> Bob: Run by the NIH
* <Joanne> FNIH Foundation for the National Institutes of Health
* <Susie> Bob: Have a large budget
* <Susie> Bob: Not much there for us
* <Susie> Bob: Not intended as integration point for multiple data sets
* <Susie> Bob: Want common data model for all data sets
* <Susie> Bob: Focus on drug outcomes
* <Susie> Bob: Want to see if claims data can be used for safety surveillance
* <Susie> Bob: Logical model cover times
* <Susie> Bob: Strong in time, causality, stats
* <Susie> Bob: Letting a thousand flowers bloom for stats
* <Susie> Bob: Have an open competition for stats
* <Susie> Bob: Have a research lab
* <Susie> Bob: Looking at the Cloud
* <Susie> Bob: Originally set up for 2 years, which ends in December
* <Joanne> Purpose is for surveillance after a drug is introduced, adverse events
* <Joanne> rather than clinical trial or translational medicine
* <Susie> Susie: Next steps regarding OMOP
* <Susie> Bob: Likely they'll want to continue after 2 years
* <Susie> Bob: Stats is more their thing
* <Susie> Susie: Colin was critical of their data model
* <Susie> Susie: Maybe through J&J I can look to see if we might be able to influence them
* <robertpowers_> Paul Stang is the J&J contact for OMOP http://omop.fnih.org/node/171
* <Susie> Bob: It's definitely a schema rather a data model
* <Susie> Bob: They think in terms of concepts
* <Susie> Bob: Then they go into a logical model
* <Susie> Bob: Occurrence is something like a bottle of medicine
* <Susie> Bob: Then model whether someone is taking their medicine
* <Susie> Bob: Can we then detect adverse events from data
* <Susie> Bob: Looking to connect eHR data too
* <Susie> Janos: Mainly relational folks
* <Susie> Janos: Mainly a relational schema for mining
* <Susie> Bob: Semantics are covered by codes
* <Susie> Bob: It's pre-coordinated
* <Susie> Bob: Believe they are beginning to hit problems with their model
* <Susie> Susie: Think we should watch
* <Susie> Chris: Important to integrate these sources of data
* <Susie> Chris: So agree keep eye on them
* <Susie> OCR
* <Susie> Ontology Clinical Research (OCRE)
* <mib_0gqank> http://hsdbwiki.org/index.php/HSDB_Collaborative_Wiki
* <Susie> Trish: Context of human studies database project
* <Susie> Trish: Main areas of focus are study design, planning for studies
* <Susie> Trish: Scanned OCRE and TMO with Samson
* <Susie> Trish: Maybe some connection in planning area
* <Susie> Trish: TMO models as a process, whereas OCRE doesn't
* <SusieS> Susie: Consider what to do in context of reviewing other ontologies
* <robertpowers_> OGMS ontology of general medical science
* <SusieS> Bob: I'll look at OGMS
* <SusieS> Trish: Richard Sherman is involved
* <SusieS> Trish: Also involved in yeast database
* <SusieS> Trish: Was meant to be the BFO for ICD
* <SusieS> Indivo Schema
* <SusieS> Michel: Spoke to EricP about Indivo
* <SusieS> Michel: Saw Chime in Toronto and discussed using CPR for modeling patient data
* <michel> http://code.google.com/p/cpr-ontology/ 
* <SusieS> Michel: Some overlap between OGMS and CPR
* <SusieS> Michel: They are both about patient records
* <SusieS> Michel: TMO doesn't have much in this area yet
* <SusieS> Michel: TMO could then map to CPR
* <SusieS> Michel: Will work to get an assessment of connection between TMO and CPR with Chime
* <SusieS> Michel: And then will work with EricP on conversion
* <SusieS> IO Informatics
* <SusieS> Elgar: No news!
* <SusieS> Susie: Abandon!
* mscottm says he's heard nothing
* <SusieS> Patient Data
* <SusieS> Scott: John Madden wanted to join us
* <SusieS> Scott: He has his hands on data through Duke
* <SusieS> Scott: Initially wanted for semantic web browser of data with snomed annotations
* <SusieS> Scott: Told him about TMO work with eHR data and he's interested
* <SusieS> Scott: He's working on deidentifying the data
* <SusieS> Scott: The records would be realistic
* <SusieS> Scott: Could also be base for creating fake data 
* <robertpowers_> Scott: Snomed annotations are not there
* <mscottm> http://bimm.stanford.edu/search
* <SusieS> Scott: Stanford may also have data
* <SusieS> Scott: It's mainly liver images
* <SusieS> Scott: Could put together demo 
* <SusieS> Scott: But it's real data
* <mscottm> http://www.himssconference.org/education/sessiondetail.aspx?eventID=3652
* <SusieS> Scott: Another option is Carolina Data Warehouse for Health
* <SusieS> Scott: Donald Spensor recently presented on it
* <mscottm> Jason Skowronski
* <SusieS> Scott: Jason told us about the data
* <mscottm> http://caties.cabig.upmc.edu/
* <SusieS> Scott: Am hoping John Madden will be able to find out more as he's in N Carolina
* <SusieS> Scott: Last option
* <SusieS> Scott: Ca ties
* <SusieS> Scott: Rebecca Crowley is running it
* <SusieS> Scott: Looking into Web site that has video, etc
* <SusieS> Scott: Best thing is to contact Rebecca
* <SusieS> Scott: Sounds fairly official, and have researched HIPPA requirements
* <jluciano> Provenance, Authoring and Versioning Ontology  (SWAN) http://purl.org/swan/1.2/pav/   Data Properties  acceptedOn createdOn ImportedOn importedFirstOn importedLastOn importedWithld lastUpdatedOn publishedOn sourceAssessedOn sourceFirstAssessedOn sourceLastAssessedOn submittedOn versionNumber   ObjectProperties  contributedBy authoredBy curatedBy contributors authors curators createdBy importedBy importedFromSource lastUpdatedBy previousVersion publi
* <SusieS> Susie: Let’s discuss provenance next time as out of time
* <SusieS> Susie: Let’s also cover Interface/eMerge next time
* <SusieS> Outreach
* <SusieS> Susie: The paper was accepted for BioOntologies
* <SusieS> Susie: Yay!
* <SusieS> Michel: Will incorporate edits I can and distribute by Friday
* <SusieS> Susie: Bio-IT Europe
* <SusieS> Susie: Seems we were too late in responding. 
* <SusieS> Susie: There will be other conferences

Additional Notes on OMOP

OMOP Observational Medical Outcomes Partnership FNIH Foundation for the National Institutes of Health

Purpose is for surveillance after a drug is introduced, adverse events, rather than clinical trial or translational medicine

Not intended as an integration point for multiple sets Instead they will make separate CDM (common data model) instance for each source data set SQL queries across all the instances

Emphasis is on drug-outcome associations across administrative domains and EHRs. They want to see if claims/EHR data can be used in surveillance mode for drug safety.

CDM Common Data Model. Person, Observation period, Drug. HOIs Health outcomes of interest: Condition, Observation, Procedure, Visit

Logical model includes considerations of time: Drug exposure, appears to be like a single bottle of medicine Drug era, appears to be a period of time when person is assumed to be on the med. Condition occurrence, Condition era Visit occurrence, Procedure occurrence Observation occurrence, Observation period

Concepts. Concept, Concept synonym, Concept relationship, Concept ancestor Seek to map each datum to a unique Concept in the Dictionary CDM should be valid even if mapping not perfectly unique. Ideally E many-to-one map from CDM data elements to Concepts in the Dictionary Everything derives from the usual set of codes, SNOMED, LOINC, ICD-9/soon10, etc. Dictionary appears to be re-coding everything (see top of p.9 of CDM) and the Dictionary appears to be completely pre-coordinated

Re-inventing the semantic wheel. "OMOP Dictionary will comprehensively address the relationships among those Concepts, such as parent- child (i.e., class-subclass) relationships, composite-component relationships" March 2010 they are even creating their own definitions of HOIs "OMOP has created, for the first time, standardized definitions of HOIs across many conditions,…" e.g. they have hired consultants to study the diagnosis of angioedema, among several other HOIs, from literature search, consistency of coding, etc.

OMOP is strong in consideration of time, causality and stats Time: pre-aggregate occurrences into Drug, Condition eras, Observation periods. Stats: approach seems to be "let a thousand flowers bloom" OMOP is running ~15 different stat methods +E OMOP Cup competition to produce 2x2 prediction/validation matrix on simulated data E simulated data: 10 years, 10M people, 4000 drugs, 5000 conditions They are focusing initial effort on 10 drug-outcome pairs, like ACE inhibitors/angioedema and warfarin/bleeding

ETL E detailed ETL specs for GE EHR and Thomson Reuters claims data --> OMOP

Challenges. 4/12/10 ppt would like, "terminology quality scores", "drug classes for high level drug outcomes", "alternative condition terminologies" There is even some comment about wanting "control populations"

Physical. There is a research lab w. ~4 people, pile of servers and a SAN, prepared to receive T-bytes of data from research partners. Not clear whether they are testing also on Amazon EC2.

Schedule. OMOP originally set up as a two-year effort. Coming to the end of phase III performance measurements July 2010: performance of methods and data in identifying drug safety issues Phase IV utility of analyses and process July-Dec 2010: effectiveness and usefulness in contributing to decision-making

Let's talk about money. Funded thru FNIH Foundation for the national institutes of health FNIH budget $10s of Ms FNIH gets Bill and Melinda Gates money to manage projects OMOP handed out money to 5 research partners