Candidate Data Sets

1
Candidate Data Sets

• Mark Thornquist
• 2005 EDRN Informatics Meeting
• 2 November 2005

2
Validation Studies

• SELDI protein patterns for prostate cancer
• Hundreds of spectra, each file a few MB
• Raw data, processed data, peaks, intensities
• DCP for hepatic cancer
• 450 patients (190 early-stage), 450 controls
• Extensive epidemiological data, single time point
  for cases, two time points for controls

3
Validation Studies

• MSA for bladder cancer
• 300 cases, 200 controls
• Panel of markers tested (15-20 microsatellite
  locations)
• Longitudinal data (up to 9 observations over 2
  years for cases)
• Intellectual property owned by for-profit company

4
Validation Studies

• Ovarian cancer biomarkers
• Joint collaboration between EDRN and ovarian
  cancer SPOREs
• Multiple markers will be tested, and a panel of
  markers is the likely outcome
• If early phases successful, marker panel will be
  tested in PLCO specimens

5
Standard Specimen Reference Sets

• Multiple sets in planning stages
• Lung, ovarian, prostate, colon, breast
• Typical design
• Rapid pre-validation and validation sets
• Several hundred specimens
• Tens to hundreds of identical sets per reference
  set
• Extensive associated epidemiological data

6
High-Dimensional Data for Discovery

• Proteomics spectra
• SELDI spectra from EVMS
• Microarrays
• 2-D gels
• Images

7
Other Categories

• Data supporting publications
• Access restricted until publication appears
• Multi-site collaborative studies
• Access restricted to collaborators until final
  release

8
Issues to be Addressed

• Human subjects requirements
• Data security
• Authorization and authentication
• Database structure
• Finding useful data subsets
• Uploading/downloading huge data sets