SV_MEETING_TITLE -- 9 Nov 2007

<matthiassamwald> impressive! ;)

<matthiassamwald> the audio signal is hardly understandable at the moment, though.

<ericP> scribe: ericP

tonya: still feel like i have to do my legacy work at Partners
... need more tool development
... tool customization is huge
... it's still easier to use legacy tools
... pain is that when i use those tools, i will need to build it again

vipul: i've not been able to articulate the long-term savings

tonya: Partners has a two year trajectory
... when we started tool investment, i had a small budget
... had to barter for resources and show resultes to keep project alive

eneumann: early discussion of ROIs
... we said "here are *possible* ROIs"
... we shouldn't try, as HCLS, to push developers
... we should instead identify a pain point that we can address in less than two years
... bridges business case

intro

eneumann: Susie learnd the outreach lessons

Joanne: decided as a group to select something, which may not align with each participants' focus

scottM: @@missed@@

eneumann: we have matured and developed a common language
... we started as an IG 'cause we didn't know what we would produce
... we can now spin off WGs producing rec-track documents

BioRDF

<jluciano> hello!

<eneumann> http://esw.w3.org/topic/HCLS/F2F/2007/11/Agenda?action=AttachFile&do=get&target=BioRDF.ppt

<mscottm> matthias - can you hear Susie ok?

-> http://esw.w3.org/topic/HCLS/F2F/2007/11/Agenda?action=AttachFile&do=get&target=BioRDF.ppt BioRDF review [Susie]

[slide 1-2]

<mscottm> I guess that means no. :)

<matthiassamwald> The audio is hardly understandable.

<eneumann> scribing... SUsie: focused more on bench side than bedside

matthiassamwald, susie moved closer to phone. i dialed in and it was ok for me

[slide 8]

<matthiassamwald> i'll try again

jar: list needs Ray Hookway

<matthiassamwald> yes, sounds better.

eneumann: KB has one mapping -- others possible
... Q: will everyone have their own mapping, or is it a shared ontology

Susie: Q: dev KB in what direction?
... .. include chem data
... .. FeDeRate data
... .. good practice for contributing

eneumann: would like to see chem data
... can invite IUPAC to join us to develop INCHI URIs
... use case: companies merging

george: [raised issue of pattented chems]

eneumann: only have to worry if police can knock on your door for writing down the chemical

alanr: 10 mins on chem during KB talk?

[yes]

Adaptice Clinical Protocols and Pathways

Adaptice Clinical Protocols and Pathways [Chimezie]

<matthiassamwald> lots of noise, i understand about 1/4 of what is said

vipul: we are discovering scenarios at Partners where we want forward-chained inference

tonya: we spend more time getting data to engine than waiting for the engine to do inference

chimezie: can do inference over a subset of the data

tonya: we are thinking about state-management
... cash inference state

vipul: that cashing required regardless of forward/backward direction

tonya: typically computer arrives at a conclusion discards them

chimezie: these are non-academic problems
... trying to see what of our needs can be captured in ontologies and what in rules

<mscottm> matthias - I'm afraid that I only moved the polycom(telcon apparatus) toward Chimezie just as he finished.

eneumann: this area had been active during the first year

vipul: inclusion/exclusion criteria is a big overlap

<matthiassamwald> there is less noise now. thanks!

chimezie: helen and i were concearned that our need for a value proposition didn't reflect exactly in this work

AdrianP: @@missed@@

<Susie> AdrianP: ACPP should collaborate with RIF.

vipul: attempt to steer OWL1.1 from our use cases leads to resisitance
... though chair (alanr) has addressed that

Susie: it's time to work with RIF

[agreement -- noted that Adrian is in RIF]

alanr: the OWL WG has a specific charter
... that said, we want use cases
... there are multiple avenues for working with OWL WG and general community
... lots of interest in temporal reasoning
... use case work involves reducing to the technical nugget

<chimezie> ACPP usecases: http://esw.w3.org/topic/HclsigDscussionTopics/HclsSubGroupACPP/Stroke

<chimezie> http://esw.w3.org/topic/HCLS/ACPPTaskForce/GuidelinesIndicationsForCABG

alanr: re chimezie's comment about what's missing, can we take sets of your [tonya's] rules

tonya: there weren't sufficient cycles to vet
... we not very proscriptive when we do decision support
... we try to make the machine notify the clinician as early as possible. can get then get negative feedback from the clinician

george: this [business rules] is low hanging fruit

vipul: we could develop and share rule archetypes

AdrianP: lots of industry vendors who would like use cases

tonya: we are working with Ilog, but found their expressivity inssuffient for our needs
... if we have 50 rules for, say, a goal lipid, we don't want to have to sequence them
... their model required attention to clinically irrelevent sequencing

[tabled]

Drug Safety and Efficacy

[slide 26 (or so) from yesterday's slides]

Bron Kissler: [introduction] was orig member of CDISC

scribe: my interests are in SDTM and where semweb works with this

eneumann: this is an XML standard. we didn't want to duplicate this work

SDTM = Study Data Tabulation Model

eneumann: non-trivial to build a large KB
... can we get *some* data and work on visualization?
... worked with Pfizer
... merged data with Similie's Babel converter
... we're still using strings here (more work to do)

[shows Ehibit of example JSON RDF]

eneumann: can, for instance, look only at "severe" event data
... data was synthesized Pfizer patient data

<alanr> http://simile.mit.edu/exhibit/

eneumann: the RDF graph has the variation in patient data and interface adds these facets automatically

<alanr> http://eneumann.org/exhibit/clinicaldemo

-> http://simile.mit.edu/exhibit/ Exhibit homepage

[eneumann demos selecting for adverse events]

<chimezie> +q

-> http://eneumann.org/exhibit/clinicaldemo Semantic Web Clinical Trials Demo

AdrianP: interface?

eneumann: Javascript

chimezie: we spent 2 hours working up an Exhibit interface. get's great wow-factor

mscottm: scalability?

eneumann: chokes around 1k records
... trying to get dev resources for this project
... Exhibit solves say 60% of interface needs as is, but can be extended with statistical functions

<Susie> Draft note on DSE at http://www.w3.org/2001/sw/hcls/Note_DSE_20071108.html

-> http://www.w3.org/2001/sw/hcls/Note_DSE_20071108 Drug Safety and Efficacy Note on CDISC's Study Data Tabulation Model (SDTM)

eneumann: interested in CDISC's extensions to model genomics

<scribe> ... new information can be added to STDM

Bron: when you get into regulatory docs, your adoption speeds up

<alanr> biont talk begins

<alanr> idea was to collaborate across the groups

<alanr> ontology best practice to be shared and driven by the other tasks

<alanr> A number of presentations were given

<alanr> developed 3 use cases

<scribe> scribe: alanr

Biont

parkinsons, combined AD use case , patient recruitment use case

COI followed from BIONT

POMR developed during ACPP. Ont group looked at this.

Helen chen's ACPP ontology

<ericP> Biont Report

discussion of difference between clinical research and clinical trials

Parkinson's disease ontology worked on at Amsterdam F2F

POMR = Problem oriented medical records

interacting with biordf was "interesting time"

<ericP> POMR wiki page

Contrast top down versus bottom up in BioRDF

Lot of momentum on BioRDF ont development, BioONT stepped back

Slide discusses a variety of issues with ontology development

<chimezie> Latest CPR ontology (was POMR ontology)

direction -> feed back to OWL 1.1 group or SWD group

Slide discusses POMR best practices

summarized by Chimezie

no "concepts" allowed.

<chimezie> Barry smith on cognitive concepts versus universals / defined classes

some discussion of issues around ontology construction

EricN SKOS versus OWL wants to be discussed

time for a break!

There is a real problem with figuring out how to represent time

scaling issues for OWL

Eric asks whether TBOX or ABOX

Next issue: How to work with ontologies and programming languages such as Java

<mscottm> :)

Vipul: interoperability - how do we make these systems work together over the next few years as both will be present

Tentative proposals: 1) Technical division 2) Application driven

for moving forward

biont and biordf need to be combined into one group

Proposed reorg: Discovery, Development, Secondary use if Healthcare Data

Alan agrees

Summary slides prepared initially by Susie, Ericp, Alan, added to by EricN

HCLS rechartering questionnaire

<jar> ericp: about w3c chartering process. community comes to w3c, proposes a charter

<jar> i.e. w3c proposes a charter.

<jar> ericp: question was how to most effectively use the energy that was out there?

<jar> ericp: we contacted many people & got many thoughtful responses

<jar> ericp: respondents were categorized, as a check on the survey

<jar> ericp: clumps - pharma, academia, vendor; w3c members/nonmembers

<jar> alanr: business development and WG spinoffs as specific activities of the IG

<jar> joanne: what is diff btwn interest group and working group?

<jar> alanr: IG scope is quite varied. could be just an irc channel; rare to do sw development as we did; but it's up to us really

<jar> ericn: this IG can be a generator [incubator] of WGs

<matthiassamwald> (for some reason, the audio quality has become worse again during the break)

<jar> matthias, the ploycom got moved away from the projector (with loud fan) and onto a table. and the speaker is standing further away from it.

<jar> ericp: join working groups are possible. e.g. w3c has worked jointly with ietf

<jar> ericp: we could do things with CDISC if we wanted to.

<jar> alanr: we can build on what we've done so far - the charter should allow us to

<jar> susie: more work with pharma space, personalized medicine (efficacy & safety)

<jar> susie: no silos please

<jar> (earlier, vipul talked about the idea of organizing vertically, by project/problem, not by kind-of-task, e.g. ontologies, uris)

<jar> ericp: there's a sweet point between thinking and typing

(shifted a while back) what should the rechartered group do & organize itself

<jar> alanr: are you still scribing, jar?

<jar> joanne: does the charter have to be done by end of day today?

<jar> alanr: a few weeks, maybe

<jar> discussion of liaisons

<jar> scottm: ...

<Zakim> ericP, you wanted to say that our current structures reflect our experiences of accountability

<jar> ericp: it has to be someone's job to reach out

<jar> susie: how does CDISC do this?

<jar> CDISC industry advisory board get to select seats on BoD

<jar> FDA is not on CDISC board, just observers

<jar> chime: liaison and advisory roles very important - need to identify stakeholders

<jar> chime: we should be pulling in people who have lots of experience e.g. Rector

<jar> vipul: how to bake some kind of outreach process into charter? [sorry vipul I'm sure I got this wrong -jar]

<jar> vipul: everyone comes for "value"

<jar> vipul: why do they not see value?

<jar> vipul: analyze this

<jar> scottm: involvement from manchester important - we don't have it as much as we'd like - carol g

<jar> scottm: how to involve more europeans and more early adopters

<jar> alanr: we should say something in the charter about this kind of activity being a task

<jar> joanne: "scout"

<Zakim> dbooth, you wanted to ask how do we make the case of what's in it for them (the early adopters) to participate

<jar> dbooth: how do we make the case to early adopters for participation in this group?

<jar> ericp: wants people to be confident that the charter will do what they want it to do

<jar> breaking for lunch now

<Susie> George's idea is that we have a Semantic Cooperation Outreach and Understanding Taskforce (SCOUT)

<jar> i'm not scribing - who is?

<ericP> did we get an introduction to Mike Travers ?

<ericP> perhaps i'll ask for that after susie's talk

<mscottm> susie: chose Alzheimer's Disease as application

<mscottm> ... lot's of different resources combined in a knowledgebase of RDF

<mscottm> ... all entities are assigned URI's and stored in RDF/OWL in a triplestore

Knowledge Base

<mscottm> ... queried using SPARQL and visualization tools for result sets

<mscottm> ... openlink virtuoso used as triplestore, HP provided hardware for server

<mscottm> alan: we're going to walk through the demo, then go into some of the ontology and data integration issues

<mscottm> ... also will look at an application that Alan built at millenium that is being rebuilt/revived

<jar> ACTION: alanr will post his presentation to wiki [recorded in http://www.w3.org/2007/11/09-hcls-minutes.html#action01]

<AdrianP> and IRC (notes) please

<mscottm> ... ask a specific question with semantic web and look at specific answers

<mscottm> ... example question: Biological processes in dendrites?

<mscottm> ... example of the SPARQL query using GO identifier and looking at label

<mscottm> ... example view of GO classes with names and definitions

<mscottm> ... possible to get same information with a GET

<mscottm> ... concentrated signal transduction and active in Pyramidal Neurons (known to be damaged in Alzheimer's)

<mscottm> concentrated ^ *on* signal transduction

<mscottm> ... web search returns too much info

<mscottm> ... interested in genes related to above signal transduction

<mscottm> Mesh: Pyramidal Neurons -> Pubmed: Journal Articles -> Entrez Gene (GeneRIFs): Genes -> GO: Signal Transduction

<mscottm> ... i.e. SPARQL query spans the above four data sources

<mscottm> ... looking at result set, many genes are related to Alzheimer's

<mscottm> ... Alan's question: create a system that will let us prioritize siRNA's provided that there are materials for studying them.

<mscottm> ... Allen Brain Atlas, mouse brain cuts stained for gene expression, 20K genes, 400K high res images

<mscottm> ... screen scraped (!)

<mscottm> ... interface built on Google Maps API to browse images

<mscottm> ... a couple of pages of code (!)

<mscottm> ... BIRN can "view source", use our code in MBAT, just like people learning by using other's html

<mscottm> ... Eric N. built a result browser in Exhibit that pulls in images and genomic context, transcription region, etc.

<mscottm> question from vipul: have you done anything with ranking results?

<mscottm> ... answer: no. That's something that text mining groups can help with.

<mscottm> question from michel: did you consider using bio2rdf?

<mscottm> answer: decided not to - wanted to be careful about the statements we were making

<mscottm> alan: shows quote from Matthias Samwald that says we haven't reached the common language yet because our query demands detailed knowledge about the heterogeneous data sources, alas: there is a lot of mapping to be done still

<mscottm> s/alan/alas/^

<mscottm> alan: about instances: "objects", particulars, independent continuants - "fully present at every time when it exists"

<ericP> vipul, using multiple ontologies manifests in differently expressed queries

<ericP> examples furnished upon request

<mscottm> ... Instances (2): "properties" (dependent continuants)

<mscottm> ... Instances (3): Processes "Takes place (unfolds) over a period of time"

<mscottm> question from Joann: do I have to be present for a walk in the woods?

<mscottm> answer from George: it depends on what plants you found along the way. <insert snare drum sound> :)

<mscottm> ... classes, roles

<mscottm> ... Words mash up functions and objects (peptide example)

<mscottm> ... Normalized representations dissect words

<chimezie> Alan Rector and Ontology Normalization

<mscottm> ... PeptideHormone defined in RDF/OWL (graph shown with sentence)

<mscottm> ... URI substituted for things in graph

<chimezie> RO - Relation ontology

<mscottm> ... 'triples' view of graph : inventory of all triples represented by graph

<mscottm> ro:has_role - from BFO

<mscottm> alan: possible answers to the "Ugly OWL RDF" problems: Macros, Rules as transformation mechanisms

<mscottm> ... Paris: Activity center analysis

<mscottm> ... analyze subgraphs based on pathways instead of single genes

<mscottm> ... background about demo: 350M triples in Openlink Virtuoso (~20Gb)

<mscottm> ... commodity hardware, biggest so far is MeSH associations to articles (200M triples)

<mscottm> ... looking at OWLIM http://www.ontotext.com/owlim/big/index.html

<chimezie> A concise argument about the danger of not separating artifacts of information systems from categories of things in reality and how it relates to healthcare terminology standards

<dbooth> Alan: What practices will best enable information to be published on the web and be widely reused and combined in new ways by people around the world? That's different from just caring about using data in a clinical setting for a more limited purpose.

<ericP> Bron: re: another CDISC touch point

<ericP> ... talking with Barry Smith

<mscottm> back a few minutes: vipul: we need to identify use cases in order to justify how we separate artifacts

<ericP> ... all orgs getting grants have to show interop with other recipients

<ericP> ... it was revealed that the commonality of interop was CDISC

<ericP> ... recipients are working out joint proposals (with CDISC) to get more grants

<ericP> joanne: Mitre is collaborating with Richard Sherman

<ericP> ... re: SWAN being users, we should record how we taught others to fish

<ericP> marcus: is there an explaination of how best to show the demo?

<ericP> eneumann: there are prepacked queries

<ericP> vipul: demo not well-enough documented

<ericP> scribe: ericP

alanr: would like offline feedback offline about diagrams et al.
... one more point: it's been a blast working with Susie et al on this demo

<jar> jar has ericn's macbook dongle

<alanr> tn: "the whole purpose of semantics is to reduce frustration"

<alanr> formula: C24H42O21 (!)

<alanr> cas:9005-79-2

<mscottm> http://bioinfo.nist.gov/SemanticWeb_pr3d/chemblast.do

<mscottm> Opening cartoon: "We'll start out by speaking in simple declarative sentences."

<mscottm> http://www.thenewyorkerstore.com/product_details.asp?mscssid=17QPRB0R3MS68L95JT1W2NDRNSTJ41JB&sitetype=1&did=4&sid=123995&pid=&keyword=cave&section=prints&title=undefined&whichpage=46&sortBy=popular

<mscottm> shorter?: http://www.thenewyorkerstore.com/product_details.asp?mscssid=17QPRB0R3MS68L95JT1W2NDRNSTJ41JB&sitetype=1&did=4&sid=123995

<eneumann> need to be clear about the info on the subject (record) or about what the record refers to

<alanr> http://www.golovchenko.org/cgi-bin/wnsearch?q=document#2n

<alanr> no sense of "about" in the way JAR talks about: http://www.golovchenko.org/cgi-bin/wnsearch?q=about

<alanr> JAR: In the future all articles will be written in RDF

<chimezie> {?ONTOLOGY definesUsageFor ?TERM. ?TERM denotes ?CLASS } => {?ONTOLOGY formalDefinitionFor ?CLASS}

<chimezie> the implied predicate concluded by that rule is not emphasized enough in this vocabulary

<chimezie> it is very important for autonomous agents to be able to find a machine-readable definition - otherwise you haven't demonstrated any value

http://view.ncbi.nlm.nih.gov/pubmed/12345

have fun

<AdrianP> There is a Pragmatic Web community which deals with pragmatics on the Web http://www.pragmaticweb.info/

- DRAFT -

SV_MEETING_TITLE

9 Nov 2007

Attendees

Contents