W3C

W3C HCLS - Clinical Decision Support

08 Mar 2012

Attendees

Present
EricP
Regrets
Chair
Matthias Samwald
Scribe
ericP

Contents


<egombocz> works - (I'm the 510-705...)

<Khan> [IPcaller.a], +1.954.633.aacc might be me

Khan: intro from Atif Khan

alan hanburry

<scribe> scribenick: ericP

alanH: colleague of matthias in vienna

<matthias_samwald> mcariaso: are you on the call now?

<cariaso_SNPedia> no, my local DNS is ailing

peterRobinson: geneticist and informaticist in Berlin, modelling human phenotypes in disease

<mscottm2> Hi Michael Cariaso! (Michael wasn't it?) We met at BioIT World Hannover 2009.

matthias_samwald: welcome to the new CDS task force
... this is part of the W3C Semantic Web for Health Care and Life Sciences Interest Group
... this is the 3rd iteration of the IG

<Joanne_Luciano> Intro: http://tw.rpi.edu/web/person/JoanneLuciano - Research Assoc Prof at Tetherless World Constellation, Rensselaer Polytechnic Institute, Troy NY USA, multidisciplinary (math, comp sci, neurosci, molecular bio, web science) multiple disease areas

matthias_samwald: we've explored how semantic technologies can be used to drive progress in basic biological research, applied pharmaceutical R&D and health care
... driving motivation is to bridge the gap between bilological research and clinical application
... in 2007, the group created one of the first large-scale RDF repo, with several hundred million triples
... we've seen uptake of these technologies in large scale systems in EU
... Translational Medicine Ontology Task Force bridged pharma research and therapies
... we made progress in mapping research to drug development leads
... the way drugs are developed is changing as it's becoming harder to bring drugs to market

<michel> Intro: Michel Dumontier - W3C HCLS Chair, Associate Professor @ Carleton University; Interest in increased understanding of medical biochemistry through biomedical knowledge representation and reasoning.

matthias_samwald: the drugs on the market are helping only a fraction of the people, and are harming a fraction as well

<mscottm2> if you don't have mute on your phone (or know how to use it), try these touchtone sequences: Mute: 61# , Unmute: 60#

matthias_samwald: "Translational Medicine" is a trendy word which is about using genetic markets to tailor therapies
... e.g. predicting adverse events to a particular compound in a particular individual
... TM hasn't been widely adopted because it's hard to juggle the hundreds of markers
... the CDS Task Force's hypothesis is that computing infrastructure will be important to helping doctors response to markers
... we are taking a step from building drug development KBs to KBs to help with drug selection or treatment choice
... we'll create prototypical implementations
... there are several ways to create decision support apps, e.g. HL7's OpenCDS which uses rules engines, it's valuable to explore how the Semantic Web stack can be more flexible and connect to large KBs

<matthias_samwald> http://www.w3.org/wiki/HCLSIG/CDS

matthias_samwald: this is summarized in the CDS wiki page <http://www.w3.org/wiki/HCLSIG/CDS>
... we have 3 listed use cases which are suggestions for task force work
... UC1 has uses LODD and patient data to propose medications
... UC2 connects SemWeb stuff to OpenCDS to provide a standard interface for CDS
... UC3 connects and enhances cancer data to create a "living review"
... that was an overview of the proposals. would like to hear how this can be connected to your projects
... we'd like a demo and a bioinformatics paper for each of the use cases

Atif Khan: at U Waterloo

<Khan> Atif Khan

scribe: constructing a CDS system which incorporates SemWeb and machine-learning techniques
... most CDS systems don't have access to all the available knowledge
... when you remove some of the information, the performance of CDS systems degrades rapidly
... just came back from IHI where we compared SemWeb and machine-learning systems

<Khan> CDC BRFSS

ericP: it sounds like you have some investment in what we want to do: DB, rules

Khan: CDC BRFSS has around 450K individuals with 450+ features
... we were looking for info which have holes in them [incomplete]
... medical records are multi-dimensional, with reports on different tests
... we worked on the 2010 edition of CDC BRFSS
... you can download it
... we needed to RDF-ize it ourselves
... we worked with a local physician to help us model a prescription protocol (e.g. if you're pregnant, don't take X) from Mayo clinic

Michael K: work on SNPpedia, incorporating 23andMe etc. to allow people to dig deaper into their genome

scribe: thanks to upstream data sources
... we're not trying to be the source or validator, just integrator

matthias_samwald: SNPedia uses semantic media wiki

<Lena> (long live the linked data geeks :D )

matthias_samwald: exposing some of the data in RDF

Michael K: it was a recognized plus, but it wasn't our decision feature

scribe: we used to have a SPARQL back end which wasn't getting much use but was doubling the memory footprint so i turned it off
... i'm on the board of semantic mediawiki foundation
... also wikidata

<cariaso_SNPedia> http://meta.wikimedia.org/wiki/Wikidata

matthias_samwald: interested in extending the RDF available?
... the goal being to integrate that data with the cloud of e.g. drug data

Michael C: afraid of a snapshot because our data is dynamic

scribe: i presume what you want is already there, unless you want extra features

matthias_samwald: some is in RDF but some isn't
... not sure this requires changes to SMW vs. just adding markup to the templates

Michael C: may be opaque to us

scribe: for each paper, what's the key gene, what are the pathways, etc, that's not avail
... we need folks to write [[pathway foo]]...
... happy to help with feature reqs, suspect that we just need more editing
... so this is a call to the community
... happy to improve templates

peterRobinson: we've developed the human phenotype ontology
... OBO ontology with about 10K isa terms
... being used by databases at NCBI and Sanger
... we've used the OWL representation
... we have a "phenomizer" product
... would love to have access to more data

<cariaso_SNPedia> minor clarifications, now that I can read the transcript (I'm on the foundation for Semantic Mediawiki . which is more correctly known as OSDA http://semantic-mediawiki.org/wiki/Open_Semantic_Data_Association )

<cariaso_SNPedia> I'm also (probably) part of the wikidata project, but there is no foundation. Its purely in the to-be-coded stage.

peterRobinson: in human genetics there are about 10 features
... mostly people type these into Google
... we've used semantic similarity searches on the ontology graph
... as a physician, it's hard to find resources for rare diseases on pubmed
... would like to say "patient with disease X, complication Y and genotype Z"
... would like to predict drug side-effects
... as genomics data increases, what's the best combo of drugs

matthias_samwald: would be great to find common ground in terms of representation and format
... michel's group is using the same technologies, with mappable ontologies

<Lena> ericP: "all standards are a popularity contest"

<Lena> ericP: irc.w3.org

matthias_samwald: the work we do here should be goal-oriented

<Khan> Two quick questions:

matthias_samwald: the goal isn't so much to have chats as to produce prototypes and document them in scientific journals

<Khan> 1. Clinical decision support target users?

<Khan> 2. What are some of the desired features of a CDS from an end user's prospective?

Khan, the COI use case was mostly patient recruitment

<Khan> ok- thanks

Khan, re: desired features of a CDS, my non-authoritative perspective is that we need to have actionable information gathered from integrative queries and inferences

matthias_samwald: CDS task force is connected to pharmacogenomics TF, both in intent and data

next telecon: 2 weeks

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.136 (CVS log)
$Date: 2012/03/08 17:12:40 $