SV_MEETING_TITLE -- 14 Jun 2012

<matthias_samwald> argh

<scribe> scribenick: ericP

Medicine Safety Code project

matthias_samwald: presented a few months ago at AMIA
... BobF and simon_lin were there
... how do we give pharmacogenomic data to clinicians without too many privacy or centralization issues?

<matthias_samwald> http://safety-code.org/

matthias_samwald: also sidestepping the sluggish uptake of EHRs
... QR-code with the ~400 SNPs of most pharmacogenomic interest
... populated by microarrays or nextgen
... QR-code points a URL
... avoiding centralization
... genetic data is associated with many privacy issues:
... .. (personal) genetic data
... .. serves as a genetic fingerprint

<matthias_samwald> http://safety-code.org/v0.1/qm70W00I2KZ0u-A01N103HmXRp0W80080zIe0oWW0vL4kC8441HTCCCfOA03U0uW020013mICe00I8wyFnk6Qo4m20e7Nq0m0500G03RG90e56407xu6O0000000000000000000000000000

matthias_samwald: law in Austria avoids genetic data

simon_lin: this URL includes the data

matthias_samwald: exactly, ~250 chars. can be carried with you

BobF: so your URL will decode the data and present the info

matthias_samwald: yes
... you can interpret such a medicine safety code can be interpreted by most phones

BobF: but your server is responsible for keeping the guidelines up to date?
... that's not a trivial task

matthias_samwald: indeed. so far just a research datatype

<BobF> ericP - the questions were from BobF (not BobF) :-)

s/boyver/BobF/g

simon_lin: given your compression, how much info can you put in?
... i assume the size of the QRcode is a design constraint

matthias_samwald: putting lots of data in a code makes it very complex and hard to scan

BobF: how many can you fit now?

matthias_samwald: i'm fitting ~400 SNPs. can't exceed ~800 with current compression

ericP: this is good for 5 years?

matthias_samwald: around 10 probably

boycer: you've got a good set of phase I/phase II
... pharmacodynamics is what your body does
... there are a hand-full of enzymes for phase I (oxidizing/reducing)
... pharmacokinetics won't change much over the next years

<matthias_samwald> http://pharmaadme.org/joomla/index.php?option=com_content&task=view&id=12&Itemid=27

boycer: but pharmacodynamics will explode over the next few years
... it's easier to predict pharmacokinetics
... but e.g. statin doesn't have a clear safe range

matthias_samwald: the 400 i use were compiled from the PharmGKB "VIP" genes (those with existing clinical evidence)
... combined with a list from pharma adne (SP?)
... added also those listed in FDA drug labels

<boycer> Matthias - what was the pharma database (I heard Pharma ad & e?)

<matthias_samwald> http://pharmaadme.org/joomla/index.php?option=com_content&task=view&id=12&Itemid=27

<boycer> Found it: www.pharmaadme.org/

matthias_samwald: pharmacodynamics would seem more interesting than pharmacokinetics, but the latter are further along
... CPIC SNPs are primarily kinetic

BobF: we're working on a schema for conveying guidelines which can be hooked into a clinical support engine
... could you use that to update safety-code.org?

<matthias_samwald> https://docs.google.com/spreadsheet/pub?hl=en_US&hl=en_US&key=0AiGT-vnkGcoLdFFVMEdqcFdYaDFqS0xHTnlUT0N3cEE&output=html

matthias_samwald: that'd be helpful
... in the last few days, i've updated this spreadsheet with CPIC and Dutch Pharmacogenomics WG guidelines from PharmaGKB
... i'd like to further curate the data to create CDS rules in SPARQL
... but it would be great the have machine-readable CPIC guidelines

BobF: that's my goal

data: application/medsafetycode+qrcode/qm70W00I2KZ0u-A01N103HmXRp0W80080zIe0oWW0vL4kC8441HTCCCfOA03U0uW020013mICe00I8wyFnk6Qo4m20e7Nq0m0500G03RG90e56407xu6O0000000000000000000000000000

ericP: use RDF/XML with controlled schema? GRDDL? conneg?

matthias_samwald: i think RDF/XML is the worst of both worlds
... the rows in the spreadsheet represent multiple SNP combos
... we can break up some of the CPIC guidelines in this table into smaller rows

BobF: i didn't intend to create a comprehensive pharmacogenomics domain model
... just serialize CPIC guidelines
... there are national efforts in the US to create generic CDS rules
... i'd like to link the schema to those rules so there's a path from the CPIC guidelines to what the hospitals can implement

ericP: how complex are these rules?

<matthias_samwald> example of guideline and boolean logic: http://pharmgkb.org/gene/PA128

BobF: most are single drug/geneotype combo
... warfarin is the most complex with a 2X2 matrix

<matthias_samwald> bite tge second row in the second table.

<matthias_samwald> note the second row in the second table

BobF: as we apply more systems bio, this will get more complex
... need to distinguish between broad clinical practics and specific guidelines
... each institution will have to decide the latter

ericP: would you want your expressivity to cover quatitative dosing even though you're not using all of that expressivity?

BobF: i'd like it to be extensible but i can't populate it now

matthias_samwald: the more complex the expressivity, the harder it is for clinicians to understand and the harder it is to validate

boycer: is the safety-code.org infrastructure going to be public?

matthias_samwald: it's a jenna back-end and i'll make the SPARQL rules open-source

simon_lin intro

UNKNOWN_SPEAKER: we're doing a trial looking at 80 genes via nextgen
... we're analyzing the impact of markers to see how we can change our practice
... marshfield clinic also has an insurance subsidiary
... we'd like to put medicine safety codes and drug interactions into an insurance card
... Mayo clinic and about 8 participants in Emerge
... expect recruitment next year

<boycer> https://www.mc.vanderbilt.edu/victr/dcc/projects/acc/index.php/About

- DRAFT -

SV_MEETING_TITLE

14 Jun 2012

Attendees

Contents

Medicine Safety Code project

simon_lin intro

Summary of Action Items

Scribe.perl diagnostic output