cabig: building a biomedical e-ecosystemsagecongress.org/presentations/buetow.pdf · • connect...
TRANSCRIPT
caBIG: Building a
Biomedical e-Ecosystem
Ken Buetow, Ph.D. Associate Director for Biomedical
Informatics and Information Technology National Cancer Institute
Sage Congress
The cancer Biomedical Informatics Grid® (caBIG®) is a virtual network of interconnected data, individuals, and organizations that redefines how research is conducted, care is provided, and patients/participants interact with the biomedical research enterprise.
caBIG®: Biomedical Information Highway
21st Century Biomedicine requires connection of diverse information
Clinical Research
Pathology Molecular Biology
Imaging
Molecular Medicine
Boundaries and Interfaces
• focus on boundaries and interfaces, how things fit together, not on the internal details • acknowledge the inherent heterogeneity of biomedical data and applications • assume that the diverse landscapes is ever changing
Research Unit
IT-enabled ecosystem!
Imaging
Microarray Data
Biospecimens
Analytical Tools
Research Center
Security Advertisement / Discovery
Federated Query Workflow
Metadata Management
Dorian GTS Index Service
Federated Query
Service
Workflow Management
Service
Vocabularies & Ontologies
GME Schema Management
Common Data Elements
Medical Center
Research Center
Research Center
Medical Center
Medical Center
Medical Unit
Research Center
Research Center
Classifying Lymphoma
• Scientific value • Use gene-expression patterns
associated with two lymphoma types to predict the type of an unknown sample.
• Connect caGrid data service (caArray) with analytical services (PreProcess, SVM and KNN from GenePattern).
• Major steps • Querying training data from
experiments stored in caArray. • Preprocessing, i.e., normalizing the
microarray data. • Predicting lymphoma type using SVM
& KNN services. • Extension
• Generalized the workflow into a cancer type prediction routine that can be used on other caArray data sets.
*Fig. from MA Shipp. Nature Medicine, 2002
Ravi Madduri Univ. Chicago
Carole Goble U. Manchester, UK
Wei Tan Univ. Chicago
Dinanath Sulakhe Univ. Chicago
Stian Soiland-Reyes U Manchester, UK
MicroArray from tumor tissue
Microarray preProcessing
Lymphoma classification
Lymphoma Prediction Workflow
Ack. Juli Klemm, Xiaopeng Bian, Rashmi Srinivasa (NCI) Jared Nedzel (MIT)
The I-SPY trial (Investigation of Serial studies to Predict Your Therapeutic Response with
Imaging And moLecular analysis):
a national study to identify biomarkers predictive of response to therapy throughout the treatment cycle for women with Stage 3 breast cancer.
(Laura Esserman, UCSF )
Multiple Morphologic Patterns of Breast Cancer
• Clinical diagnosis
• Treatment history
• Histologic diagnosis
• Pathologic status
• Tissue anatomic site
• Surgical history
• Gene expression
• Chromosomal copy
number
• Loss of heterozygosity
• Methylation patterns
• miRNA expression
• DNA sequence
Specialized Programs of Excellence (SPOREs)
Multiple Sites/ Organizations
Multiple Data Types
Cancer and Leukemia Group B (CALGB)
American College of Radiology Imaging Network (ACRIN)
University of California at San Francisco (UCSF)
I-SPY Trial: Identify biomarkers predictive of therapeutic response in Stage 3 breast cancer
10
I-SPY Adaptive Trial Outline
Accrual: Anticipate 800 patients over 3–4 years
Enroll ~20 patients per month
Participating Sites: 15–20 across US and Canada
On Study
MRI MRI Biopsy Blood
MRI Blood
Surgery
Biopsy Blood
MRI Blood
Tissue
Taxol +/–New Drug (12 weekly cycles)
AC (4 cycles)
Taxol + Trastuzumab* + New Agent A
Taxol + New Agent C
Taxol + Trastuzumab*
Taxol + Trastuzumab* + New Agent B
Taxol
AC
AC HER 2 (+)
HER 2 (–)
Randomize
Randomize
Surgery Taxol + New Agent D *Or Equivalent
On Study Surgery
Taxol + Trastuzumab* + New Agent C
Surgery
Learn, Adapt from each patient
Taxol + Trastuzumab* + New Agent A
Taxol + New Agent C
Taxol + Trastuzumab*
Taxol + Trastuzumab* + New Agent B
Taxol
AC
AC HER 2 (+)
HER 2 (–)
Randomize
Randomize
Surgery
Taxol + New Agent F
Taxol + New Agent D
Taxol + New Agent G *Or Equivalent
Learn, Adapt from each patient
On Study Surgery
Taxol + Trastuzumab* + New Agent C
Taxol + Trastuzumab* + New Agent F
Surgery
I-SPY Trial IT Infrastructure
Expression Array Data!SNPArray Data! Radiological Data!Clinical Data!Patient Samples!
Data Mart
API
Local System
caExchange - Hub
Tolven
API caBIG® Applications
caTissue caArray
API
Perspectives for building a Community where disease data and models are shared
Lessons Learned: • Culture eats Strategy for Lunch: Most of what prevents data-sharing is not technical, but reflects the segmentation of stakeholders into silos (by discipline, by geography, by sector), or concerns about intellectual capital, getting credit
• Begin with a “coalition of the willing” • There are no free lunches
• Failure to collect appropriate structured data can result in loss of information or inability to connect
• context is critical for clinical data • Exchange AND use of information requires more than the ability to electronically move it from point to point. • Pay me now or pay me later
Challenges Ahead: • The culture of science is driven towards doing “new” not leveraging the work of others and incentives are generally aligned that way – “not invented here”. We don’t move forward because we are always starting over.