Sponsoring Cancer Center

1
Integrative Cancer Research

• Sponsoring Cancer Center
• Lombardi Comprehensive Cancer Center
• Georgetown University
• Workspaces
• Architecture (developer)
• Integrative Cancer Research (developer)
• LCCC caBIG Representative
• Robert Clarke, Ph.D., D.Sc.
• clarker_at_georgetown.edu
• [phone]

2
Integrated Cancer Research Teams

• Georgetown University
• Edmund Gehan (Biostatistics Biomathematics)
• Stephen Moore (Advanced Research Computing)
• Seong Ki Mun (Imaging Science Information
  Systems)
• Cathy Wu (Protein Information Resource)
• Virginia Tech
• Joseph Wang (Engineering Computer Science)
• Catholic University
• Jason Xuan (Engineering Computer Science)
• National Institutes of Health
• Spiderweb Team (NCICB )
• Javed Khan (NHGRI)
• Aiyi Liu (NICHD)
• University of Edinburgh, Scotland
• William Miller (Oncology)

3
Overview

• Project Activity
• Tool development for exploring very high
  dimensional data sets
• Deployment of grid enabled and integrated array
  (MIAME
• compliant) and clinical research databases
  (Spiderweb)
• Contribution of expression array and other data
  from multiple
• clinical, translational, and basic research
  projects
• Stage of Maturity
• Tools at each stage of development
• Technical Details/Standards
• Open source currently most are written in C
  and/or MatLab
• Points of Interoperability
• Use of caCORE APIs caBIG-compatible APIs
• Resources
• Platform migration software engineering
  personnel

4
Examples of Ongoing Funded Projects

• Bioengineering Research Partnership (NCI)
• Large, prospective, molecular profiling study in
  breast cancer
• Expression array tool development and deployment
• Expression array and clinical data
• Gene/Nutrition Interactions in Breast Cancer
  (NCI)
• U54 program project array tool development and
  deployment data
• Breast Cancer Center of Excellence (DOD)
• Study of alcohol and breast cancer risk array
  tool deployment data
• Clinical Translational Research Study (DOD)
• Large, prospective and retrospective, molecular
  profiling study
• of Letrozole and Tamoxifen in breast cancer
• Array tool development and deployment data

5
Examples of Ongoing Funded Projects

• Computational Decomposition of Composite
  Molecular
• Signatures (NCBIB)
• Expression array tool development, optimization,
  and deployment
• Intelligent Mapping and Visual Exploration of
  Gene
• Expression Profiles (NCI)
• Expression array tool development, optimization,
  and deployment
• Comprehensive Computational Analysis of Gene
  Expression
• (NCI)
• Expression array tool development, optimization,
  and deployment

6
Integrated Tool Development
7
Tools for Very High Dimensional Data Sets

• General Approach
• Probablistic approaches to address the
  properties of very high
• dimensional data spaces (e.g., "curse of
  dimensionality",
• "concentration of measure phenomenon")
• Data Preprocessing
• Normalization
• Tissue heterogeneity correction
• Multitask, goal-specific, gene selection tool
  (e.g., classification vs.
• signaling/function)
• Cluster Discovery and Visualization
• Visual Statistical Data Analysis (VISDA) package
• Classification and Prediction
• Optimized multilayer perceptron classifiers
• Expression array tool development

8
Comprehensive Data Analysis
Cluster Discovery Visualization
Data Preprocessing
Classification Prediction
Adaptive Hierarchical Subspace Experts
Cross-Phenotype Normalization
Tissue Heterogeneity Correction
Information Visualization
Optimized Mutlilayer Perceptron Classifiers
Visual Statistical Data Analyzer (VISDA)
Multitask Gene Selection
Matlab
Neural Networks
Pattern Recognition
Independent Component Analysis
9
Examples of Novel Tools
Discriminant Components Analysis
10
Dynamic Contrast Enhanced-MRI
Tumor angiogenesis in the breast
11
Integrating Imaging and Molecular Analysis
12
Integrate DataProtein Information Resource
13
SpiderWeb Collaboration Goals

• Integration of multiple, varied information to
  achieve a basis for rational design of novel
  diagnostics and therapeutics
• Integration of functional genomics information
  into clinical information so it can be used to
  diagnose genetic predisposition,
  sub-classification of disease and help with
  optimal selection of therapies
• Bring new efficiencies to clinical research by
  integrating bench to bedside and back

14
SpiderWeb Project (LCCC/NCICB)
Approval
15
Timelines and Resources

• Development Requirements (approx)
• 3 programmers for data analysis tools
• 3 programmers for PIR integration into caBIG
• Infrastructure
• C/Java development environment enhancements
• software system administrator
• Draft 12-month Work Plan and Milestones (approx)
  
• 1-3 months VISDA caBIG interoperability
• 4-6 months CPN and THC caBIG interoperability
• 7-9 months MTGS and MLP/AHSE caBIG
  interoperability
• 10-12 months software system integration and
  final testing
• 12 months for PIR integration into caBIG