HCLSIG/PharmaOntology/Meetings/2009-09-24 Conference Call

From W3C Wiki

Conference Details

  • Date of Call: Thursday September 24 2009
  • Time of Call: 11:00am - 1pm 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 instructions: go to http://www.mibbit.com/chat and click the server link. Enter irc.w3.org:6665 into that box, enter a nickname, and enter #HCLS2 for the channel
  • Duration: 2h
  • Convener: Susie

Agenda

  • Select the use case to implement - all
  • Review of ontology - Colin
  • Namespace for the ontology - all
  • Maintenance of the ontology - all
  • Application demo - Jun
  • F2F meeting - Susie

Minutes

In attendance: Susie, Tim, Jun, Colin, Christi, Julia, Joanne, Elgar, Michel, Chris, Chimezie, Matthias

Apologies: Bosse, Anja

<colin> susie: christi, chris and susie have done work on use case

<colin> susie: then colin will give overview of ontology

<colin> susie: then discussions re namespace, best practices & maintenance

<colin> susie: jun will demo application based on linked data and neuroscience

<colin> susie: and we'll discuss f2f if we have time

<colin> susie: topic - Use cases

<colin> http://esw.w3.org/topic/HCLSIG/PharmaOntology/UseCases

<colin> susie: not a linearly written use case

<colin> susie: built patient-centric use case with separate branch based on the research role in a pharma company

<colin> susie: real rather than contrived even if it makes our job harder

<colin> susie: scenario - 1. patient goes to physician with symptoms (different in alzheimer's)

<colin> susie: 2. physician makes diagnosis and gets test done. snp profile?

<colin> susie: 3. now physician goes to application with his data incl. medical history, concurrent medication, lifestyle information.

<colin> susie: application makes recommendations

<colin> 4. then maybe brain scan is performed and we have clear diagnosis

<colin> 5. physician goes to application for selecting drug and dosage based on patient data - mass, snps, concurrent medication

<colin> 6. or scenario where physician prescribes common drug - nasty side effects so the physician prescribes different drug with different side effects, etc.

<colin> 7. now imagine clinical trial - check whether patient fits inclusion/exclusion criteria and which arm to go into

<colin> susie: then move onto research scenario

<colin> susie: related use case

<colin> susie: 1. Medical information from patients signed up for the Varenicline? trial is sent to AD researcher in pharma to further investigate the genetic underpinnings of the disease

<colin> susie: 2. Researcher compares medical (SNP) data from patients to normal – and confirms APOE appears to be a significant factor – nothing new here..

<colin> susie: 3. New interesting research comes in from a GWAS study in Nature genetics implicating CLU, and PICALM

<colin> susie: 4. Virginia Lee presents at a conference that TDP-43 might be important

<colin> susie: 5. Investigates gene->protein-> pathway & MOA 6. See if pathway has known side effects

<colin> susie: 7. Identify known druggable targets in the pathway # 8. Look at all compounds known to interact with that target (extends to compound families and fragments) # 9. Explore whether existing drugs (on market or in trials) interact with the newly implicated pathways # 10. Discover drug in trial is implicated in pathway

<colin> susie: # 11. Go back to patients enrolled in trial to determine prevalence of SNP across arms – perhaps recruit further patient to a new arm # 12. ….Or there wasn’t a compound, and start hunting for leads.

<colin> susie: v interested in feedback

<colin> chris: most important to me that there's a single e-portfolio for the patient accessible by anyone (IP, privacy permitting)

<colin> chris: and to which new information can be added

<colin> chris: also, what is the toxicity data?

<colin> chris: and also contains information about current trials

<colin> chris: way ahead of it's time, of course...

<colin> susie: international repercussions - v fragmented in usa due to payer structure, but more unified in uk, wouldn't capture clinical/snp/lifestyle data

<colin> susie: mature vendors in the united states hard to displace

<colin> joanne: patients who move country

<colin> susie: indivo developed by Ken Mandl (Harvard) is code for personal health records somewhat adopted by google & microsoft - DOSSIER consortium adopted by some employers - also snp data and lifestyle data e g from Garmin

<colin> susie: also agreements about amount of data you want to share; different levels of consent

<colin> susie: AND pharmas are buying data off payers

<colin> susie: and also people exploring payment methods

<colin> susie: maybe micropayments for selling ehrs

<colin> susie: interested in other people's input on business models etc.

<colin> julia: primary concern about immediate future and scope

<colin> julia: simpler use case -- too much here

<colin> susie: identify subcomponent

<colin> julia: maybe pick two or three questions and go from there

<colin> susie: Christi - what thoughts?

<colin> christi: I'd like to take a smaller subset as a proof of concept

<colin> joanne: should keep big picture

<colin> susie: key deliverable is the ontology; persuade vendors to use this?

<colin> susie: next steps - go through ontology and use cases side by side

<colin> julia: make things look bigger by linking the item in the ontology to other public sources - parallel work stream

<colin> susie: capture both where the use case links into our ontology and also other data available

<colin> susie: formalize definitions and originating ontology - do this in owl or separately

<colin> colin: done! but also mention in the paper

<colin> susie: we should identify subset by next week's call

<colin> susie: but how?

<colin> susie: can you help us understand the data sources used by a physician in the patient scenario

<colin> susie - chris?

<colin> susie: colin - focus on ontology work

<colin> susie: susie to work on research scenario

<colin> susie: michel - plug tmo terms into both use cases

<colin> susie: trish and joanne - go through both use cases and identify links out to other ontologies

<colin> susie: elgar - work with colin/trish

<colin> trish: would like to work with colin on annotations

<colin> chris: what about the insurance sector?

<colin> susie: vipul kashyap will be happy to look at the ontology

<Julia_Kozlovsky> Susie: last week we talked over use cases, and not too much time about the ontology

<Julia_Kozlovsky> Susie: how will we synch all the ontology versions. Latest on the wiki page? Other options?

<Julia_Kozlovsky> Susie: is there a collaborative Protege tool?

<Julia_Kozlovsky> Colin: no version control in the above?

<Julia_Kozlovsky> Trish: that's correct.

<Susie> The ontology is at http://esw.w3.org/topic/HCLSIG/PharmaOntology/Ontology

<Julia_Kozlovsky> Colin: Google code?

<Julia_Kozlovsky> Colin: sourceforge does not like if actual code is missing - Google code doesn't care..

<Julia_Kozlovsky> Colin: took the spreadsheet, added terms from external ontologies where exist, created ids, put names into RDFS labels and descriptions into comments

<Julia_Kozlovsky> Colin: many terms are now subclasses of other public ontologies (information artifacts..)

<Julia_Kozlovsky> Colin: some don't have parents yet..

<Julia_Kozlovsky> Colin: efficacy, safety, lifestyle - definitions will depend on data we use

<Julia_Kozlovsky> Colin: for biomarker, we can define a parent once we see how data is used now..

<Julia_Kozlovsky> Susie: are there any open q's to discuss?

<Julia_Kozlovsky> Colin: there are 'don't know' or 'need more work' sections

<Julia_Kozlovsky> Colin: what data examples will we have for those?

<matthias_samwald> isn't toxicity a disposition?

<Julia_Kozlovsky> Colin: Protege 4 works well..

<michel> quantifiable entities include: efficacy, comparative effectiveness, toxicity, dosage

<Julia_Kozlovsky> Is there a special AutoID plugin?

<michel> biomarker is a role

<Julia_Kozlovsky> Colin: yes, we should use it..

<Julia_Kozlovsky> Joanne: there are errors when loading

<Julia_Kozlovsky> Colin: P4 shows labels rather than IDs

<Julia_Kozlovsky> Colin: let's fill out definitions for efficacy and so on..

<Julia_Kozlovsky> Colin: in XML, the list is on the bottom..

<Julia_Kozlovsky> Trish: let's put this into curation status - add a new property

<Julia_Kozlovsky> Susie: can Elgar take the lead on defining the terms and pointing to available data.

<Julia_Kozlovsky> Susie: which namespace to use? We can use W3 URL, or Trish can provide one?

<Julia_Kozlovsky> Susie: are there implications from URL to ownership? What matters?

<Julia_Kozlovsky> Trish: OBO foundry ontologies have terms referenced by purl.org.. Not sure what the policy is.

<Julia_Kozlovsky> Colin: waiting for a response on OBO policies.

<Julia_Kozlovsky> Susie: unless we hear otherwise from OBO, why don't we go with W3C namespace?

<Julia_Kozlovsky> Trish: purl seems to be where standards are going..

<chimezie> it probably comes down to what behavior you want when you fetch the NS uri. If all you want is just the OWL, then it seems like a wash either way. If you want something more sophisticated (documentation page, for instance), then it depends on what can be done on either host (purl.org v. w3c.org)

<Julia_Kozlovsky> Michel: use document URL as base, with '#' added for each term - to be changed later

<Julia_Kozlovsky> Joanne: uncurated, ready for release, placeholder, pending final vetting, metadata incomplete, metadata complete, example to be eventually removed - list for curation status

<Julia_Kozlovsky> Colin: are we happy using Information Artifact ontology?

<Julia_Kozlovsky> Trish, Joanne, Susie: fine

<michel> Fine with using IAO

<Julia_Kozlovsky> Michel: Chimezie depends on what we want when we fetch the URI..

<trish> here are a few examples that ncbo is setting up

<trish> - http://purl.bioontology.org/ontology/NPO.owl resolves to an OWL file, so people can use it for imports, etc. - http://purl.bioontology.org/ontology/NPO takes you to the visualization of the root node in BioPortal - http://purl.bioontology.org/ontology/NPO/Nanoparticle (or any other concept) takes you to that concept in BioPortal

<Julia_Kozlovsky> Colin: fetch should resolve to an HTML page?

<Julia_Kozlovsky> Susie: if we use W3C, we can decide what happens..

<jluciano> This may be useful for us - http://obi-ontology.org/page/OBI_Minimal_metadata

<Julia_Kozlovsky> Susie: will keep talking to W3C

<Julia_Kozlovsky> Susie: in the paper, we need to address maintenance

<Julia_Kozlovsky> Susie: will need edits over time?

<jluciano> Mandatory metadata: MUST be provided for both classes and relations

<Julia_Kozlovsky> Trish: it gives history and helps with de-duplication..

<Julia_Kozlovsky> Joanne: the 'should' vs. the 'must'?

<jluciano> uncurated Nothing done yet beyond assigning a unique class ID and proposing a preferred term

<Julia_Kozlovsky> Trish: we should definitely get in the mandatory ones and change the properties a bit to conform

<jluciano> metadata_incomplete Class is being worked on; however, the metadata (including definition) are not complete or sufficiently clear to the branch editors.

<jluciano> metadata_complete Class has all its metadata, but is either not guaranteed to be in its final location in the asserted IS_A hierarchy or refers to another class that is not complete.

<Julia_Kozlovsky> Christi: what is our mandate? How far should we go?

<jluciano> pending_final_vetting All definitions, placement in the asserted IS_A hierarchy and required minimal metadata are complete. The class is awaiting a final review by someone other than the definition editor.

<Julia_Kozlovsky> Susie: undefined, but we did not explicitly talk about maintenance..

<jluciano> ready_for_release Class has undergone final review, is ready for use, and will be included in the next OBI release.

<Julia_Kozlovsky> Susie: hoping not much is required.. Perhaps there should be a mailing list to respond to comments and make change decisions?

<Julia_Kozlovsky> Colin: Google code provides a mailing list capability. The scope is of concern - somebody has to mind the results..

<Julia_Kozlovsky> Christi: maybe a subset of this group can take this on?

<Julia_Kozlovsky> Susie: there can be extensions built by other groups

<Julia_Kozlovsky> Consensus: maintenance - yes

<Julia_Kozlovsky> Susie: Jun's application (www.open-biomed.org.uk) is a linked data app that uses Anja's data and can be an example for our use cases

<Julia_Kozlovsky> Jun: this app is disease and med - centric, rather than patient-centric

<Julia_Kozlovsky> Jun: demonstrating the advantages of linked data

<Julia_Kozlovsky> Jun: Alzheimer's is used as example

<junzhao> http://www.open-biomed.org.uk/admed/admedapps/searchTCMByDiseaseName

<Julia_Kozlovsky> Jun: search for Alzheimer's disease, and then click on 'Ginko Biloba'

<Julia_Kozlovsky> Jun: from each herb, links are to web sources are on left, and the effects are from linked data sources..

<Julia_Kozlovsky> Susie: let Jun go on with the demo, but the rest can wrap up..