Break out sessions

1
Break out sessions
2
1. What standards are being used to describe
Genetic Variation Data2. What infrastructure
resources are available/what is needed for data
integration3. How should research interact with
the health service4. What funding model is
appropriate5. What should the NCRI be doing in
this area
Aims of the meeting
3
Workshop Format

• Invited speakers representing the breadth of the
  NCRI portfolio
• Breakout sessions in 3 groups with questions
• We have drafted some already
• We are happy to take these from the floor too
• Report back

4
Breakout groups
Group 2 Rakesh Nagarajan Richard Wooster Aengus
Stewart Andrew Devereau Gillian Heap Tim
French Angela Cox Stephan Feller Jay Kola Peter
Kerr
Group 3 Paul Lewis Crispin Miller Wendy
Russell Adrian Moody Cecilia Lai Heike
Grabsch Andy Bush Dawn Smith Paul Flicek Helen
Parkinson

• Group 1
• Anthony Brookes
• Tina Boussard
• Norman Freshney
• Jane Cope
• Chris Mattocks
• Audrey Petitjean
• Richard Begent
• Margus Lukk
• Liz Pittendreigh
• Daniel Zicha
• Max Wilkinson

5
1. What standards are being used to describe
Genetic Variation Data2. What infrastructure
resources are available/what is needed for data
integration3. How should research interact with
the health service4. What funding model is
appropriate5. What should the NCRI be doing in
this area
Aims of the meeting
6
Group 1-Q1 Standards

• Discussion
• Standards and their role
• Reporting semantics
• Role of Journals
• Community buy in and uptake
• Conclusions
• Essential to drive Standards development
• Semantics fundamental
• Best standards evolve naturally (flexible)
• Role of Journals and Stakeholders
• Potentially biotech/industry

7
Group 1 Q2 Resources

• Discussion
• Ontologies
• Individualisation
• Harmonisation
• Changing concept of data bases
• Funding
• Conclusions
• Domain is of great importance
• Ability to annotate either locally/centrally
• Community driven consortia
• Novel database models (self-sustaining)
• High dimensional data challenges
• Cancer provides ideal model

8
Group 1 Q 5 What can NCRI do?

• Discussion/Conclusions
• Forums and facilitation
• Communication to funders/stakeholders
• International relationships
• Consensus building
• Expanding workshop portfolio (series)

9
Infrastructure

• caBIG strongly semantically typed
• PML not semantically-typed will PML-2 support
  an ontology
• Should NCRI mandate full caBIG-compatibilty re
  standards
• Much existing data that doesnt meet caBIG
  standards NCRI funders should be sympathetic to
  this
• Other caBIG infra caCORE, caDSR
• caGRID have demonstrated joining across
  different data types by
• UK myGRID workbench for researchers for using
  bioinformatics resources (ongoing work)
• HL7 SIG genotype/pedigree family object
  modelling project
• Choosing the right tool for the job e.g. NGRL
  are developing website that lists resources/tools
  - which ones are better than others, how well
  are they curated, quality of data etc
• Is it important to standardise algorithms?
  Perhaps not as long as raw data is available
• Any tool is potentially useful (unless its out
  of date)
• Tricky researchers/end users often want yes/no
  answer but quite often an interpretation
• Some metrics might be useful about updates,
  usage, etc

10
Infrastructure(2)

• Should funders insist that tools have some
  quality metric e.g. a developer carries out a
  control test on the tool they have developed
• Generalise to other areas
• Balance between directed research (e.g.
  NCRI-compatible) and solely cutting-edge
• Moving from research to development means funding
  dries up
• Also cultural issue in that service-provision is
  different than research researchers dont
  necessarily want to provide it
• OPEN-SOURCE v PROPRIETARY
• Both are necessary which is best model/balance.
  open-source okay if support is in place
• caBIG is welcoming companies to buy-in to
  standards and possibly to provide support
• Sanger has mixture e.g. heavy use of Oracle but
  Ensembl runs off mySQL
• Industry AZ must run off certain platform which
  has advantages within company but tricky sharing
  data outside
• People might be initially reticent about
  uniformity of platform but often end-up happy
  with the benefit

11
Infrastructure(3)

• Mass negotiation by NCRI on behalf of research
  community for commercial software could be very
  economically beneficial
• Agreement that if funders are paying for the
  development then software should be open-source
  but support could be proprietary
• Infrastructure for SHARING data as opposed to
  ANALYSING data should always be open-source
• OMG standard no-one using it

12
CFH

• No family ID because patient-centred how do we
  exchange pedigree data? Confidentiality issues
  about sharing this type of data
• Is SNOMED-CT good enough to handle genetic data?
• Is HL7 good enough for exchanging genetic data?
• Can slow things down because these are very big
  standards organisations
• Do Once and Share project looking at Clinical
  Genetics (Andrew Devereau involved)
• Difficult to ask people for consent when being
  diagnosed
• Secondary uses services can export pseudonymised
  data for research
• Research phenotype data is much more diverse
  than CFH phenotype data therefore exchange
  standard far away
• (Are there established standards for ethnicity?)

13
What NCRI can do

• Maybe working groups can be established to adress
  issues
• Involvement with other initiatives lobbying
• EU holding a workshop in March (standards and
  funding)
• HGVS (re standards)
• tighter alignment with caBIG workspaces
  (Integrated Cancer Research workspace most
  suitable)
• NCI/NIH funding?
• Industry would benefit and should be approached
  (e.g. like SNP consortium)
• NHS Informatics (funding?)
• Patient advocacy groups are very powerful also
  important to give feedback to patients (important
  to have a good public face)
• Important to be able to sell the benefits to
  organisations like CFH and industry

14
Group 1

• Discussion/Conclusions
• raw data is needed for gen var (HTP)
• ontologies are a problem
• phenotype covers everything incl. Histology
  slides
• Clinical definitions complex
• Mapping to genome is less of a problem than the
  phenotype
• Mapping only 70 pc of the data is OK the rest
  may not be good enough
• Protocols vary also need standardisation of the
  processing as well as the reporting
• This costs money, clinicians dont want to do
  this as std
• Funders have the power
• Training sets might help e.g for badly sampled
  tissues vs useful ones
• Rare diseases find stdizn easier data
  sharing/data integration
• Meta analysis is a way forward, id whats been
  reported most and work from there
• Conclusions
• 1
• 2
• 3

15
Pt 2.

• Bioinformaticians suffer as part of MDTs non
  bioinformaticians managing informatics projects
  can be problematics
• Transatlantic cooperation is desirable, interest
  in the caBIG process
• Training is key