Steve Harris, CancerGrid talk - 10 March 2006

At the moment research done via paper mostly. Protocol and data capture forms etc.

Aims

• Open standards for breast cancer clinical trials Enhanced data discovery Routine meta-analysis Data ingegration

Methods

• Harmonised vocabularies Different data stores
  • Need semantics of measurements recorded somewhere

Benefits

• Tamoxifen (drug given to prevent breast cancer recurrence) Herceptin VIOXX Mostly down to intelligent interpretation of trial results That's what "meta-analysis" means - computers helping with that process.
  • Always done in the past. Done on paper. Laborious.
  Want to embed software in people's work (even if that means plug-ins for MS Word etc.etc. Open standards, open source, free at point of use

Networks

• Virtual collaboration support, video conferencing, identity management Integration, security, workflow, access Need to delegate - boss delegating the typing in of notes to other clinician etc. Identity/roles etc. an issue so that access to data is affected. Methods
  • Clinical trials portal Message level security

Things that others could reuse

• Personal video conferencing (not access grid!) ontologies and standards inside user tools XML databases with full MDG Web services that talk to both Java and .Net

Why did they drop personal access grid?

• Too hard for clinicians Have been using something called Polycom PVX
  • Standards based H.323

  More deployable than personal AccessGrid Software about £90 (return ticket to London!?!)

Protocol designer

• Click a button in MS Word that builds an XML description of the trial.

eXist

• Free open source XML database stack
  • Schema binding etc. etc.
  Almost there in terms of XML database being able to draw clear UML diagrams

Gaps

• Distributed integrated security (Shibboleth?) Semantic web - topic maps/RDF/OWL (OWL ontology too new, unfinished)

Cancergrid in context (its place in e-Health at Oxford)

• CGs push for open documented standards in the NHS (e.g. how do they measure blood pressure and record it)

  Push to think about data rather than 'systems'

  • Really, the only thing that matters is "data". The system can be built around it. Can have multiple systems.
    • e.g. we've been measuring blood pressure for decades (hundreds?) of years, but our understanding of what it means has changed hugely.
  Trying to push idea of semantic grid
  • On the desktops In default working environment