Joel Saltz

1
(No Transcript)
2

• Joel Saltz
• Chair and Professor
• Biomedical Informatics Department
• Professor Computer and Information Science,
  Pathology
• The Ohio State University

3
Overview

• Grid based clinical research infrastructure
• OSU Grid Software Infrastructure
• CALGB and cardiac clinical research studies

4
Grid based clinical trials supportWorldwide
Scope for Clinical Research Studies
1000s of potential clinical research sites,
different studies involve different subsets of
sites Different sites can use different names for
the same entity Semantic grid, SNOMED,
LOINC Support for authentication, encryption,
anonymization, role based data access Support for
grid data aggregation Grid based coordination of
clinical studies Must leverage pre-existing
medical IS systems each of these are complex
trigger based federated systems
5
Clinical Research Grid Types of Information

• Radiological Studies
• Pathology
• Molecular (Proteomics, gene expression)
• Genetic, Epigenetic (SNPs, haplotype analysis)
• Laboratory, pharmacy, outcome data

6
Aggregation of Data in Virtual Information
Warehouses
Virtual Information Warehouses
Tissue bank Lab Data
Clinical Genomic Data
Clinical Data
Clinical Data
7
Clinical Research GridMore than just data
aggregation

• Define iteratively define clinical protocols
• Changes arise from scientific, institutional
  review and with ongoing analysis of study data
• Patient accrual
• Identify suitable patients, obtain patients
  consent for study
• Execution of protocol
• Maintain and execute rule base in order to carry
  out treatment and testing specified by protocol
• Ongoing assessment of patient data determines
  patient treatment, what data will be obtained,
  which specimens will be collected, how the
  specimens will be processed, which tests will be
  carried out
• Patient safety and protocol optimization
• Ongoing analysis of data from overall study and
  of data from individual patient

8
Analysis Prediction of patient outcome,
effectiveness of treatment relationship of
genomic data to pathophysiological measurements,
outcome
Drives accrual, protocol changes, choice of
laboratory, imaging, genomic testing
Data streamed to Analysis
Analysis subscribes for data updates
Workflow Execution of rule-based protocols,
execution of algorithms that specify tests and
treatments, coordination of patient consenting,
specimen collection and analysis
Generates requests for data
Data Diagnosis, Treatment, Laboratory, Imaging,
Proteomic, Gene Expression, Gene Sequence
Data driven algorithms -patient accrual,
clinical, laboratory, genomic testing
9
Overview of Clinical Research Grid

• Customized access control
• Institutions and patients decide what data to
  share
• Ad-hoc data warehouses
• Each research project and consortium can maintain
  its own data view
• Institutional databases linked to grid
• Grid based molecular dataset and image analysis
• Images as first class objects

OSU Information Warehouse
10
Clinical Research Environment at Single Site OSU
IS Infrastructure
11
Components of Local Information System

• Electronic medical record
• Clinician order entry, clinical protocol
  specification and tracking
• Laboratory System
• Digital Radiology (PACS)
• Datagate triggers invoked by message monitoring
• Appointments, billing
• Logistics scheduling people and resources
• Information warehouse
• Security, single sign-on

12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
Infrastructure for Clinical Trials Support at OSU
Layered on OSU Information Warehouse
17
CPR Decision SupportFacilitate Best Practice
through POE
Order sets designed to support evidence based
clinical guidelines
Standard templates/defaults for complex orders
18
Ohio Clinical Trial Research Consortium
19
(No Transcript)
20
Federate Emerging Databases
Infrastructure to relate, combine produce
meta data
Deformation Segmentation Quantification
Slide courtesy of Arthur Toga
21
Two groups have developed BIRN project
partnerships, so far

• Mouse BIRN - Animal Models of Disease / Multi
  Scale/Multi Method - MS Mouse and DAT KOM (a
  schizophrenic and otherwise interesting mouse
  animal model)
• Brain Morphology BIRN - Targets neuroanatomical
  correlates of neuropsychiatric illness (Unipolar
  Depression, mild Alzheimer's Disease (AD), mild
  cognitive impairment (MCI)

22
OSU Grid Software Infrastructure

• Support for optimized query, processing, analysis
  of distributed datasets
• Integration of software with NSF PACI software
  suite (Globus, SRB, Network Weather Service)
• Collaboration with BIRN

23
Software Support for Data Driven Applications

• DataCutter Component Framework for Combined
  Task/Data Parallelism
• Filtering/Program coupling Service Distributed
  C component framework
• GridDB Lite Large Data Query Layered on
  DataCutter
• Indexing Multilevel hierarchical indexes based
  on R-tree indexing method.
• Data Cluster/Decluster/Range Query
• Active Proxy G Active Semantic Data Cache
• Employ user semantics to cache and retrieve data
• Store and reuse results of computations

24
DataCutter

• Flow control between components
• Schedulers place filters on grid processors
  (scheduler API)
• Stream based communication
• MetaChaos data descriptor, data mapping support
  for inter-component data transfers
• Data aggregation implemented as a component
• NPACkage

Download at www.datacutter.org
25
Integrating DataCutter with existing Grid
toolkits SRB, Globus, NWS

• SRB integration Subset and filter datasets
• Globus integration DataCutter uses Globus
  resource discovery, resource allocation,
  authentication, and authorization services.
• Network Weather Service (NWS) integration NWS
  for used for system monitoring.

Distributed by NPACI as NPACKage
26
GridDB LiteSelect Operation on Grid Data
?Distributed Array
27
Query Planning
Query
Data Source
Filter
Index
Distribution Generation Service
Data Source
Filter
Distributed Program
28
Query Execution
Data Mover Service
Distribution Generation Service
Partition
partition
Data Mover Service
Distributed Program
29
Multi-Query OptimizationActive Proxy G
q1

• Goal minimize the total cost of processing a
  series of queries by creating an optimized access
  plan for the entire sequence Kang, Dietz, and
  Bhargava
• Approach minimize the total cost of processing a
  series of queries through data and computation
  reuse
• IPDPS2002,SC2002,ICS02

q2
This blue slab is the same as in q1
We have seen the pieces of q3 computed for other
queries in the past
q3
30
Grid Based Image Analysis Toolkit

• Framework to support distributed image processing
  applications
• Use DataCutter, VTK, and ITK
• Provide a standard framework for describing image
  processing workflow and its data in order to
  enable creation of image processing Grid Services.

• DataCutter Distributed workflow system used for
  building applications that can operate in a
  cluster computing environment.
• VTK The visualization toolkit used for creating
  visualization applications of all kinds which
  will manipulate and view image data.
• ITK Insight segmentation and registration
  toolkit is quickly becoming the standard toolkit
  for the archival and invention of image analysis
  algorithms.

31
NPACI Telescience, BIRN and MicroscopySupport
for Telescience Portal using VTK, ITK
40,000 pixels

• Goal
• Remote access to and processing of subsets of
  large, distributed images.
• Even single images can be very large (a few
  hundred MB to tens of GB per image for montaged
  digitized microscopy images).
• Support by DataCutter for
• Basic database operations Indexing, querying,
  and subsetting of large images and image
  datasets.
• Image processing supported by VTK (Visualization
  Toolkit) and ITK (Insight Segmentation and
  Registration Toolkit) layered on DataCutter.
• Use of heterogeneous, distributed clusters for
  data processing.
• DataCutter is part of NPACKage which also
  includes SRB, Globus, and Network Weather Service
  as an integrated suite of tools.

40,000 pixels
Query
DataCutter
Telescience Portal
32
Telescience Portal and DataCutterDemo at NPACI
2003 All Hands Meeting (March 03)
Compute Cluster
Storage Systems
Globus
Storage Resource Broker (SRB)
DataCutter Filter
VTK/ITK
DataCutter
Globus
Compute Cluster
Globus

• Middleware Tools
• DataCutter -- subsetting, filtering, and
  processing of data in a distributed environment
• Globus -- Authentication, resource allocation,
  and remote process execution.
• SRB -- file I/O to different storage systems.

• With DataCutter
• Some of the processing can be done near data
  sources to reduce volume of data.
• Compute intensive operations can be executed on
  collections of compute clusters.

33
Radiology Clinical Studies using Dynamic
Contrast Imaging

• 1000s of dynamic image sets per clinical study
• Iterative investigation of image quantification,
  image registration and image normalization
  techniques
• Assess techniques ability to correctly
  characterize anatomy and pathophysiology
• Biopsy results
• Changes in tumor structure and activity over time
  with treatment
• Images from many sites including NIH, Heidelberg,
  Oklahoma, Ohio State
• Collaboration with Michael Knopp, MD

34
(No Transcript)
35
Large Scale DataCutter Testbed Analysis of Oil
Reservoir Simulation Data

• Evaluate geologic uncertainty and production
  strategies simultaneously
• Multiple realizations of multiple geostatistical
  models
• Multiple production strategies (number, location
  of wells)
• Dataset Size 5TB
• 500 simulations, selected from several
  Geostatistics models and well patterns
• Each simulation is 10GB
• 2,000 time steps, 65K grid elements, 8 scalars
  3 vectors 17 variables
• Stored at
• SDSC HPSS and 30TB Storage Area Network System
• UMD 9TB disks on 50 nodes PIII-650, 128MB,
  Switched Ethernet
• OSU 7.2TB disks on 24 nodes PIII-900, 512MB,
  Switched Ethernet
• Data Analysis
• Economic model assessment
• Bypassed oil regions
• Representative Realization Selection for more
  simulations

36
Clinical Studies
37
Leukemia Correlative ScienceCommittee (LCSC)

• Michael A. Caligiuri, MD
• Cancer and Leukemia Group B

38
CALGB Leukemia Tissue Bank (80,000 Vials)
Bioinformatics
39
LCS Committee Administration

• Michael A. Caligiuri, MD (Chair, 1999-)
• John Byrd, MD (Vice Chair, 1999-)
• Thomas Look, MD (Vice Chair, 2000-)
• Stephen George, PhD (Faculty Statistician)
• Jennifer Shoemaker, PhD (Faculty Statistician)

40
CALGB LCSC History

• 27th year of existence, substantial impact
• Focus molecular classification and molecular
  targeting
• For last 18 years, the LCSC grant has been
  submitted as a separate U10 rather than as part
  of the Chairs grant
• Critical for initiating and implementing a strong
  laboratory program in correlative science
• Greater flexibility in managing the reference
  laboratories
• Provides direct access and control of funds by
  the scientists
• Allows rigorous peer review of laboratory science
  and guarantees a base of support for the LCSC
• Alleviates additional administrative strain on
  the Central Office

41
CALGB LCSC Purpose

• Pursue the biologic basis for malignant
  transformation in hematopoietic cells
• To identify the cytogenetic, cellular, and
  molecular aberrations of transformation
• To understand the relationship of these processes
  to clinical outcome in order to significantly
  increase the fraction of patients cured of
  leukemia

42
Summary of LCSC Activity

• Eight LCSC protocols open
• Eleven Leukemia Tissue Bank studies opened
• Six LCSC Protocols completed or closed
• Eight LCSC Protocols in development
• 4,391 accruals to LCSC Protocols ( 71 ? )
• 41 Peer-reviewed manuscripts
• 43 Peer-reviewed abstracts presented
• 14 Primary peer-reviewed manuscripts submitted

43
LCSC Overview of Major Accomplishments

• LCSC identified distinct groups of leukemia
  patients whose cytogenetic or molecular
  characteristics proved to be predictive of
  clinical outcome
• Implemented clinical trials whose treatment
  paradigms included stratification based on the
  presence or absence of markers identified as
  relevant to prognosis by the LCSC

44
The Hemizygous FLT3 ITD Genotype Predicts Poor
Prognosis in AML Patients lt 60 Years of Age with
Normal Cytogenetics
Time (Years)
Similar results Thiede et al. Blood
994326-4335, 2002
45
LCSC Research Themes and Aims

• Novel Molecular Markers in AML (Dr. Gilliland)
• To prospectively evaluate the prognostic
  significance of
• the hemizygous Flt3 genotype
• BAALC gene expression
• predictive value of additional novel molecular
  markers following preliminary analyses

9621 9720 19808 10102
46
LCSC Research Themes and Aims

• Molecular Profiling in Leukemia (Dr. Staudt)
• Develop a genetic and/or epigenetic expression
  profile or signature in AML, T-ALL, B-ALL, and
  CLL cases that lack specific indicators of
  clinical outcome relevance to clinical outcome.

9420 9621 9720 19808 10101 10103
47
(No Transcript)
48
What are the best predictors of outcome?
CALGB LTB Provision of Patient Material
Clinical
Correlation
Statistical Ccnter Bioinformatics Unit
Microarrays For An Integrative Genomics
Storage QC/QA Validation Algorithms
49
Pharmacogenetics-Pharmacogenomics

• OSU Program in Pharmacogenomics
• Director W. Sadee
• Focus on genetics of complex diseases
• and therapy
• Cardiovascular
• CNS
• Cancer
• Chemogenomics

50
(No Transcript)
51
Treatment of Coronary Artery Disease How to
Exploit Genetic Information for Optimizing
Therapy

• Atherosclerotic process components
• Cholesterol metabolism and transport
• Inflammation
• Coagulation
• Drug response
• Statins

Glen Cooke, Heart Lung Research Institute Joel
Saltz, Biomedical Informatics Heifeng Wu,
Coagulation Clinics Bo Yuan, Genomics Wolfgang
Sadee, Audrey Papp, Julia Pinsonneault,
Pharmacology Clay Marsh, Heart Lung Research
Institute Xiaotong Shen and Dan Dougherty,
Biomathematical Sciences Institute
52

• Overall Approach CAD study
• Large number of candidate genes
• Haplotype multiple phased sequence variants
• (SNPs) obtain genotype
• Gene dosage
• 4. Allele-specific mRNA analysis
• 5. Associations with clinical phenotype
• In vitro analysis of proteome, transcriptome
• (plasma, monocytes, plaque tissue)
• Functional assays (e.g., monocytes)

53
Cholesterol Lowering Therapy with Statins
Prediction of therapeutic efficacy Pravastatin,
simvastatin, lovastatin Inhibitors of HMG-CoA
synthase
Kuivenhoven et al. - variant of the cholesteryl
ester transfer protein gene in the progression
of coronary atherosclerosis. NEJM
199833886-93 intronic common SNP function?
54
CETP Genomic Structure

• Taq1A/B in linkage disequilibrium with promoter
  polymorphisms
• I5045V frequent SNP in coding region
• (marker SNP for allele-specific mRNA analysis)
• Exon 9 splice variant lacking exon 9 yields
  dominant negative CETP
• Approach measure specifically CETP and exon9-
  mRNA,
• each by allele-specific assays
• (allele-specific PCR or allele-specific ligation
  PCR)

55
CETP LinkageDisequilibrium
56
Cholesterylester transfer protein gene (CETP)
OSU CAD study (Glen Cooke) 950 patients and 500
controls
Possible number of haplotypes 2416
57
Associations with Risk Factors
58
Major Adverse Coronary Events
59
Lipid Profiles comparinggenotypes
60
Collaborators OSU College of Medicine Glen
Cooke Joel Saltz Catalin Barbacioru Dan
Cowden William Abraham UCSF Plasma
Membrane Transporter Group Vera Rhakmanova Ed
Bilsky Gordon L. Amidon Patsy Babbitt Carmine
Coscia John Weinstein Kimberly Bussey
Acknowledgements
Sadee Lab Pascale Anderle Jung-eun Lim Danxin
Wang Julie Lucas Julia Pinsonneault Audrey
Papp Xiaochun Sun Ying Huang Ying Zhang Daniel
Dougherty Zunyan Dai
61
Thanks to

• State of Ohio BRTT
• National Institutes of Health
• National Science Foundation
• OSU Comprehensive Cancer Center
• Department of Energy
• DARPA