HCLSIG/PharmaOntology/Meetings/2010-09-02 Conference Call

From W3C Wiki

Conference Details

* Date of Call: Thursday September 2 2010 
* Time of Call: 12:00pm - 1:00pm ET 
* Dial-In #: +1.617.761.6200 (Cambridge, MA) 
* Dial-In #: +33.4.26.46.79.03 (Paris, France) 
* Dial-In #: +44.203.318.0479 (London, 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

  • Next Steps/Strategy Discussion - Susie
  • TMO Updates - Michel, Elgar
  • Patient Mapping - Michel, Chime, EricP
  • Interface/eMerge - Bosse, Chris
  • Outreach (UPenn Translational Medicine, AMIA 2011 Summit on Translational Bioinformatics, etc.) - All
  • AOB


Minutes

  • Attendees: Bob, Trish, Matthias, Joanne, Elgar, Chime, Bosse, Christi, Michel, Scott, Eric, Susie
  • Apologies: Chris

<bobP> Susie: Want to focus on next work

<bobP> ...2nd want to focus on patient data, in an Indivo-way

<bobP> ...Propose to start w current work, then strategy

<bobP> ...Michel gave a good talk, now TMO invited to another conference

<bobP> ...Polish for the general paper

<bobP> ...Patient-centric data: Eric, Indivo

<bobP> Eric: Mixed optimism. Start w. demographics, then patient records were largely generated at 3AM

<bobP> ...not formally worked out w. ontology.

<bobP> ...Indivo is interesting b/c open-source schema, have been feeding them

<bobP> ...they have a bag of concept codes snomed, makes sparql queries awkward and data are largely unreachable

<bobP> ...How much to we normalize the data for blood pressure, etc?

<bobP> ...w. Ben discussing an rdf representation

<bobP> ...Have some XML --> Indivo --> RDF

<bobP> ...Exists person, w. tests; what does CPR capture w. this? Does it capture all of this?

<bobP> Chime: CPR not terminology-oriented.

<bobP> ...exercise to get better ideas of how to work

<bobP> Eric: Cool! want to steal that!

<bobP> Q: on ontology?

<mscottm> Q is Scott:

<mscottm> Use an ontology of units for measurements?

<bobP> Eric: patient encounter is easy(?) w sparql query

<mscottm> Work has already been done on such an ontology

<bobP> ...Yes, we will use units, want to normalize before we use it.

<bobP> ...mm Hg vs bar conflict: 120 to 2

<bobP> ...When cherry picking in first order form, that same xslt will perform unit conversions

<bobP> Susie: Michel, you outlined list of patient-centric target projects

<bobP> Michel: Queries need to involve patients, so we should spend time reviewing Eric's work

<bobP> ...Also, need to better understand how the data are being generated

<ericP> -> http://esw.w3.org/HCLSIG/LODD/Data LODD Datasets

<bobP> Susie: Not sure how data are being updated

<ericP> http://esw.w3.org/HCLSIG/LODD/Data has documented update cycles, but i have no idea how that happens

<bobP> ...NCT updated about once a quarter

<bobP> Inclusion/exclusion criteria should be incorporated explicitly

<bobP> Q Do you use drugbank? There may be some issues w. drugbank that have been encountered.

<bobP> ...should be able to get from RDF graph when it was updated, other context.

<bobP> Mappings between datasets might have to be regenerated; we need a pipeline to keep data fresh.

<bobP> Susie: Anja publishes in Berlin, but problem downstream w. keeping data fresh.

<bobP> ...maybe host at W3C

<bobP> We cannot depend on other people to do the scripts; we need copies.

<bobP> Scott: Other alternative, we have some datasets, maybe we can keep them up to date and functional

<bobP> Susie: Maybe call w. Anja's associate re governance of data

<bobP> ...would help w. 3 of the datasets that we need

<bobP> Susie: We have clear next steps for LODD datasets

<bobP> Scott: Need wiki page improvements, we don't have many queries where they actually amalgamate

<bobP> ...need to look at our queries, formulate using TMO terms and relations, not specific to e.g. diseaseome set

<bobP> ...We can query across datasets w our own TMO terms. We are too focused on the dataset in the existing queries.

<bobP> Susie: Michel, useful to use tools to frame thinking?

<bobP> Michel: Student wrote script to querie RDF graph, (I guess a meta-view?)

<bobP> ...show me a path from this type to that type

<bobP> ...Will get student to run his tool on our TMO data, share next week.

<bobP> Susie: Eric connecting w. Chime to understand how to use CPR/Indivo

<bobP> Chime: Curious about general value of using a specific ontology. What are terms to represent in high fidelity.

<bobP> Susie: Eric and Chime over the next week can talk about queries

<bobP> Michel: Once these projects are settled, we'll be in good shape

<bobP> ...need to make sure the reviewers are happy. Start Oct 1 on paper.

<bobP> Suie: J Biomed Sem is an open journal. Downside is we have to pay $ to publisher.

<bobP> ...how are we doing w. contributions?

<bobP> A: several instances of $200 contributions: We have $1000 + Joanne $150 then we have the cash.

<bobP> Susie: Strategy and Next Steps for TMO

<bobP> ...Asked Michel to be lead for TMO going forward. Passing the torch from Susie to Michel!

<bobP> Thank you, Michel.

<bobP> Hail to our new leader! Thank you Susie!

<mscottm> Thank you Susie!

<bobP> Susie: would be interested in TMO going back to apps, like identification of targets

<bobP> ...or identification of key mechanisms of disease, those sorts of areas. As opposed to interface for EMR system.

<bobP> Michel: Mechanism of disease is very interesting.

<bobP> ...Descriptions are subtle and important, and would show value, to show circumstances and context.

<ericP> does AlzForum use any NLP in their curation?

<bobP> Bosse: Agree w. Susie and Michel. Need to keep link between biomed and patient data.

<bobP> Michel: Patient data focus was from the aspect of the clinical trials researcher

<bobP> ...can we find another drug w/o some side effect. Really nice use case in clinical trials

<bobP> Joanne: Can we do something w. respect to outcomes. If we can take something w effect on outcomes

<bobP> ...then this shows something that really makes a difference.

<bobP> Michel; One of our queries is what drug can patient take under Medicare D?

<bobP> Joanne: Some drugs have higher toxicity, can person tolerate a different drug but w/o same tox for this patient

<bobP> Michel: Organize by chem structure, organization?

<bobP> Joanne: Yes. We need impact, like safety and efficacy. If we can keep targeted on these kinds of priorities then these would be good use cases.

<bobP> Michel: Curated pharma code on depression, pharmacogenomics or pharmacogenetics?

<bobP> Susie: Important to identify biomarkers

<bobP> Michel: Lots of info not available. There are tools to help curate articles in order to build a database.

<ericP> can we centrifuge a few scientists and test for differential expression of ideas?

<bobP> Joanne: Shaw(?) at EPA might have data on liver tox

<bobP> Michel: Datasets that are available and that are structured are low hanging fruit.

<bobP> ...we may have to get to data that other can't get to yet, w tools. Might be too much.

<bobP> Susie: Can we find structured datasources that are low-profile, then get data from unstructured text.

<bobP> Michel: On HCLS wiki we spent some time on use cases (which are quite cool by the way)

<bobP> ...let's look again at those queries. Animal models, for example, not much data available.

<bobP> ...we may have to pull these data together ourselves

<bobP> ...Mechanisms of disease, pathways, core bio-inform stuff. But also animal model.

<bobP> Joanne: This is a very exciting discussion.

<bobP> Michel: Next week come back to this discussion. What to work on. Look at TMO use cases and your own interests.

<mscottm> and congrats on getting the baton for TMO!