Experiences from CancerGrid and caBIG

1
Experiences from CancerGrid and caBIG

• what are the real issues in international
  interoperability?
• Peter Maccallum
• CancerGrid Project Manager
• Department of Oncology
• University of Cambridge

2
Overview

• Some National Grids for Cancer Research
• UK CancerGrid
• US caBIG
• Technical interoperability
• Security infrastructure
• Data integration infrastructure
• Semantic interoperability
• Shared terminologies and standards
• Shared models

3
CancerGrid

• CancerGrid is a consortium of leading researchers
  from the UK cancer and computing science
  communities developing open standards for
  clinical cancer informatics.

Department of Oncology, University of Cambridge
Weatherall Institute of Molecular Medicine,
University of Oxford Division of Cancer Studies,
University of Birmingham Department of Oncology,
University College London Centre for Cancer
Research and Cell Biology, Queens University
Belfast Oxford University Computing
Laboratory Cambridge e-Science Centre Midlands
e-Science Centre Belfast e-Science Centre
4
Phase III breast cancer trials
                                                 
                   
 
Neo-tAnGo
Eligibility Criteria (i)Histologically confirmed
invasive breast cancer (ii)T2 tumour and
above (iii)Diameter gt20mm
Arm A CMF Classical x 6 cycles
Arm B E x 4 cycles / CMF x 4 cycles
SAMPLE OF BLOOD TO BE COLLECTED ON ONE OCCASION
ONLY FROM THE ABOVE TRIALS AT FOLLOW-UP OR DURING
CHEMOTHERAPY  
SNP PROFILE GENOTYPE ANALYSIS CORRELATION WITH
CLINICAL TOXICITIES AND OUTCOME
5
caBIG

• cancer Biomedical Informatics Grid
• An initiative of the NCI in the USA, linking
  hundreds of cancer reserarchers with a shared
  technology stack.
• Key technologies
• The Enterprise Vocabulary Service (EVS)
• The Cancer Data Standards Repository (caDSR)
• Grid technology (caGRID)
• Biomedical object models (caBIO)
• Clinical trials object models (BRIDG)

6
Technical issues

• Certificates issued by caBIG CA.
• No recognition of other CAs (e.g. UK eScience).
• Overall access to the Grid is under US control.
• Fine-grained access control must be added by
  providers.

7
Technical solutions

• A peer-to-peer trust network.
• Role-based access control to provide access to
  some shared resources.
• Clear, legally enforced agreements on role
  membership criteria

8
Current status

• Adding UK nodes to caGRID
• caGRID 0.5 is based on GT 3
• Weve added 4 nodes to the caGRID 0.5 network
• The nodes trust each other, and share services
  with the US
• Future technical developments
• caGRID 1.0 will allow us to register web services
  directly via GT4

9
Data integration issues
caBIG Vocabulary

• Data sources are wrapped (using hibernate).
• Resources conform to caBIO models.
• Models are registered in the caBIG data standards
  repository.
• Data standards reference NCI vocabularies.

caBIG Data Standards
caBIG Models
US
UK
10
Data integration solutions
Vocabulary
Vocabulary

• Data sources conform to models.
• Resources conform to standard models where
  available.
• Models are registered in local data standards
  repositories.
• Data standards reference validated vocabularies

Data Standards
Data Standards
Models
Models
US
UK
11
Current status

• Working with caBIG
• We currently use the caDSR as our data standards
  repository via web services
• Weve evaluated some caBIG software and models
  directly (caTissue, caArray)
• Building a compatible infrastructure
• Were deploying vocabulary services based on
  caBIG technology (LexGrid from the Mayo clinic)
• Weve built a lightweight, caDSR compatible data
  standards repository

12
towards interoperable Grids?
Knowldge
Knowldge
caBIG Semantics
CancerGrid Semantics
Standards
Standards
Models
Models
US
UK
caBIG Technology
CancerGrid Technology
Trust
13
Overview

• Linking Software Grids
• Linking with caGRID is technically feasible
• True interoperability will depend on shared
  security infrastructures
• Linking Semantic Grids
• Linking models and messages is where the real
  long term interoperability will be found
• If we can prove the benefits, many more cancer
  research networks could be linked through shared
  models.

14
Acknowledgements

• Technical interoperability
• PV Jithesh (QUB), Ioannis Daskalopoulos
  (Cambridge)
• Oxford eSC, Cambridge eSC, Midlands eSC
• George Komatsoulis (NCICB)
• Semantic interoperability
• Steve Harris (Oxford), Igor Toujilov (UCL), Maria
  Lin (Birmingham), Andrew Tsui (Oxford)
• Harold Solbrig (Mayo), James Buntrock (Mayo)

15
Funding
The Medical Research Council provides support
through grant G0300648 (2005/06 to 2008/06) for
12 full time staff, 1 graduate student, travel
and equipment.
The EPSRC provides support through grant
EP/C531272/1 (2005/05 to 2008/05) for travel and
meetings.
Microsoft Research provides support for 2
graduate students, software and equipment.