See also: IRC log
scribenick Bob
<AmitSheth> there is nothing at http://www.w3.org/wiki/HCLSIG/SWANSIOC/Actions/RhetoricalStructure/
Michel: Use case plus the data
and the ontological discussions
... paper is about SNPs pxgx, how SNP's might be recorded in
patient records
<joanne> pls post link to paper again. i see only a link to rhetorical structure
scribe: breast cancer case: info
still tied around gene expression as a biomarker
... a little bit further away from what we wanted to embark
<joanne> 12:09 Bob https://docs.google.com/document/d/1lKdDSb2uBBIeTEQAv2CyTHN_aVW63k9si1hmmilOMi0/edit?hl=en&authkey=CJGUtcwF#
<joanne> for bo
Michel: Breast cancer case suited for collab w BioRDF
<AmitSheth> +q
Michel: maybe a separate paper. One paper for SNPs, one paper for pharmacogenomics
Joanne: TMO to bridge;
Michel: Introduction by Eiker(?),
startup w breast cancer interest
... there is a front end that uses TMO; based on what we
published
... Learned as much as there is about warfarin
... dosing, SNP ids; also prediciton algorithms, interesting to
see lots of work
<AmitSheth> was this shared on this group? http://news.sciencemag.org/sciencenow/2011/04/computer-algorithm-may-speed-dru.html?ref=hp
Michel: pharmGKB is interested. Wonder if we can look at semantic web services
<AmitSheth> +q
Michel: would be well placed to do prediction for a speicific patient
Elgar: What are predictions based on?
Michel: Testing is done by int'l
normalized ratio INR: how active is the pathway for
coagulation
... amt of warfarin needs to be balanced
... there are home monitoring kits to manage dosing
... factors other than the two genes are diet, etc that can
change
... this is already a nice patient-specific case
Joanne: bc w BioRDF; warf
w...
... Lib of Personalized Med has published on warfarin
Michel: Also genetic tests for
low $
... genetic component explains 40% of dosage variance
... FDA has published the table reproduced in the doc
... but the table is less effective than other research
Elgar: We have two things that
determine dosing
... so there is an input vector for the algorithms
... dosage recommendation is not great for accuracy
<joanne> what are the inputs?
Elgar: we would like to look at 1) can we find other features to consider? and 2) look at diff prediction algorithms
Michel: There are papers that
crit pgx approach; doesn't work for people >65
... reduced metabolism, loss of liver function, etc
<AmitSheth> Looks like being on the queue does not help: so here are questions/comments: (a) have you thought about the quality of data you are integrating? (b) quality of intergration?
Michel: consider a range of factors beyond genetics, should be able to account for more variance
(well, warfariin dosing is tough!)
Elgar: No idea at moment how data analysis and prediction would sound
Michel: Have touched the surface; maybe take closer look at prediction aspect
Joanne: Interested in prediction, along w Elgar
Michel: We need a better
understanding of what they use to train
... one of papers in section on prediction; performance of the
different approaches
... are there some elements that are being missed, even in
state-of-the-art
... could we add more features to explain more variance?
<joanne> here's one tonellato warfarin avatar paper: http://www.google.com/url?sa=t&source=web&cd=4&sqi=2&ved=0CDAQFjAD&url=http%3A%2F%2Fpeople.dbmi.columbia.edu%2F~rip7002%2FSite%2FHome_files%2FSimulated%2520Comparison%2520of%2520Warfarin%2520Treatment%2520Protocols%2520Chi%25202010.pdf&ei=itDCTeG-K9PqgQeMlJHaBg&usg=AFQjCNGRNZdX7CCrLw3Mh5YJK3pIHkV7Yg&sig2=GmqL31HAsPP
What is quality of data that you are going to use?
Michel: Reduce dimensionality of
data; can analytically find strong correlates in the
input
... we would complement what they already use w new
contributions
<joanne> Look at page 17 and 18 of this (at least!) http://docs.google.com/viewer?a=v&q=cache:gaOqmO8SKj4J:online.law.asu.edu/events/Personalized_Medicine/ppts/Monday/Afternoon/Session_III/Tonellato.pdf+tonellato+warfarin+filetype:PDF&hl=en&gl=us&pid=bl&srcid=ADGEESi-OP2QSvCcWEh-8WEj1Ukq0ApfCCv3_xqKfTNFBWbTdKlZ3_TzyCVp475MFh1xs1qbV8YTcHVXGWnfxpXHSE8HXiak2sH59WHNMhSzd
Amit(?): question about correlates and alignment
Michel: Can't comment on even
their quality of data
... won't build a new algorithm for paper
... how do we represent facts; mostly focus on rep of
knowledge
... we might build a better predictor later :-)
... Looked at pathways from pharmGKB; basically a diagram
... don't know how well maps out to Reactome or Kegg
... There was a GWAS on pharmGKB; want to convert patient data
wrt outcomes
<michel> http://www.pharmgkb.org/do/serve?objId=PA135603152#tabview=tab2
that's not the right link!
here's another one:
<michel> http://www.pharmgkb.org/do/serve?objId=PA165291561&objCls=PhenotypeDataset#tabview=tab1
Michel: Based on clinical and
other demo factors
... this is the most important dataset for us; need to convert
to semantic set for individual
... convert to RDF; ensure that our representation has all the
fields, and in ontologies
... ontological nature of the representations
Joanne: will do tab-to-RDF
Michel: Are there other datasets that we might consider?
Joanne: emailing warfarin paper
to everyone
... can set parameters flexibility in the avatars
Michel: Which field; does our ont
have the slots?
... vis-a-vis predictive stuff that Russ did? Which fields
might be missing, overlap of these.
Joanne: Probably have captured all the important ones
Michel: Warfarin case: enough for good investigation
Bosse: Would like to see link to pharam co
Michel: Label has been updated
twice; gene but no dosing
... 2009 updated w table, w variance accounted for
... What kind of data are required to make decision? expensive
to do trials
Bosse: Clinical trial could be the angle here.
Michel: Also what data the FDA requires so that becomes part of the label?
Amit: Interested to see how this
is realized. Issues of data access, quality etc
... have developed tools to produce RDF
... capturing data is what I would like
Matthias: Interested in clinical
decision support; how pgx data can be used for alerts etc
... seamless integration into clincial workloads
Michel: Adrian Coulet and I will work on representations
This is scribe.perl Revision: 1.135 of Date: 2009/03/02 03:52:20 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) No ScribeNick specified. Guessing ScribeNick: Bob Inferring Scribes: Bob WARNING: No "Topic:" lines found. WARNING: No "Present: ... " found! Possibly Present: Amit AmitSheth Bosse Elgar Joanne Matthias Michel bbalsa epichler hcls2 https iker joined You can indicate people for the Present list like this: <dbooth> Present: dbooth jonathan mary <dbooth> Present+ amy WARNING: No meeting title found! You should specify the meeting title like this: <dbooth> Meeting: Weekly Baking Club Meeting WARNING: No meeting chair found! You should specify the meeting chair like this: <dbooth> Chair: dbooth Got date from IRC log name: 05 May 2011 Guessing minutes URL: http://www.w3.org/2011/05/05-hcls2-minutes.html People with action items: WARNING: No "Topic: ..." lines found! Resulting HTML may have an empty (invalid) <ol>...</ol>. Explanation: "Topic: ..." lines are used to indicate the start of new discussion topics or agenda items, such as: <dbooth> Topic: Review of Amy's report[End of scribe.perl diagnostic output]