ASSESSING
THE REAL RISK
IN
COMPLEX DISEASES
Michael N. Liebman, PhD
Chief Scientific Officer
Windber Research Institute
Overview
Data, Information and Knowledge
 Systems Biology
 Defining Translational Research
 Understanding the Question(s)
 Clinical Breast Care Project (CBCP)
 Windber Research Institute
 Data Integration

Gap
AMOUNT
DATA
INFORMATION
GAP
GAP
GAP
KNOWLEDGE
KNOWLEDGE
CLINICAL UTILITY
TIME
Systems Biology
(Personalized Medicine)
Patient
Physiology
Genomics Proteomics
Metabolomics
CGH
-omics
Bottom Up Approach
Physiology
Genomics Proteomics
Metabolomics
CGH
????
Top Down Approach
(Personalized Disease)
Patient
Physiology
Genomics Proteomics
Metabolomics
CGH
Translational Medicine
Basic
Research
“Bench”
Training
Job Function
“Language”
Culture
Responsibilities
Clinical
Practice
“Bedside”
Translational Medicine
“Crossing the Quality Chasm”
Closing
The Gap
Clinical
Practice
“Bedside”
Basic
Research
“Bench”
Training
Job Function
“Language”
Culture
Responsibilities
Humans as Detectors

Characteristics
–
–
–
–
–
–
–
Spectral sensitivity (visible region)
Sound sensitivity (audible range and volume)
Memory (retention is critical for comparison)
Perception (focus on what is known)
Analytical Capability (simple vs complex)
Ranks Importance of Change by Size (Bias)
Evolves slowly compared to other technological
advances
– Does not perform uniformly over 24/7
“Discovery consists in seeing what
Everyone else has seen and thinking
What no one else has thought”
A. Szent-Gyorgi
Asking the Right Question is
95% of the Way towards
Solving the Right Problem
Defining a Patient

A 48 year old woman, married, 2 children
(ages 18, 24), presents with an abnormal
mammogram, biopsy shows presence of
cancer which, upon extraction, is diagnosed
as invasive ductal carcinoma (T3,M1,N1).
Her2/neu testing is +2
1. Modeling Disease
Disease as a State vs Disease as a Process
 Bias of Perspective
 Temporal Perspective

Modeling Disease
{ Risk(s)}
{Disease(s)}
Lifestyle + Environment = F(t)
| Genotype |
(SNP’s, Expression Data)
Phenotype
(Clinical History and Data)
|
UMLS Semantic Network
??
Disease Etiology
DIAGNOSIS
Genetic
Risk
Lifestyle
Factors
Breast
Survival
Cancer (Chronic Disease)
Pathway of Disease
Natural History of Disease
Treatment History
Outcomes
Treatment
Options
Environment
+ Lifestyle
Disease
Staging
Patient
Stratification
Early
Detection
Genetic
Risk
Biomarkers
Quality
Of Life
Her2/neu (FISH) = Her2/neu (IHC)
Her2/neu (IHC1) = Her2/neu(IHC2)
Do Either Measure the Functional Form
of Her2/neu?
Phenotype
Childhood Diseases
| Genotype | |
Smoking
Menarche
Overweight
Diabetes
Cardiovascular Disease
2nd Hand Smoke
Natural History ?
Breast Cancer
(Age 48)
Phenotype
TIME
Longitudinal Interactions
in Breast Cancer




Identify Environmental Factors
Quantify Exposure
– When ?
– How Long ?
– How Much ?
Extract Dosing Model
Compare with Stages of Biological
Development
Lifestyle Factors
Smoking
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
AGE
Obesity
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
AGE
Alcohol
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
AGE
2. Genetics and Disease

Genetic Pre-Disposition
– < 10 % of all breast cancers
– Not all BRCA1 and BRCA2 mutations
result in breast cancer
- Modifier genes?
- Lifestyle or environmental factors?
- Pedigree Analysis
Pedigree (modified)
Influenza Pandemic 1918
1940
DES
Time
1950
1960
1970
Measles
Polio Vaccine
Influenza
Influenza
1980
1990
2000
Menopause
Prostate Cancer
PSA
3. Aging and Disease
Processes of Aging vs Disease Processes
 Ongoing Breast Development
 Same Disease : Different Host?
 Text Data-mining Approaches

Disease vs Aging
Hormone
Replacement
Menarche
Menopause
Perimenopause
Child-bearing
<50 years>
Heart Disease
Breast Cancer
Ovarian Cancer
Osteoporosis
{
{
Alzheimer’s
Aging
Disease
Quality of Life
Breast Development
Cumulative
Development
Lactation
Menopause
Menarche
Peri-menopause
Child-bearing
Ontology: Breast Development
Parous
Terminal Buds
Buds
Lobes
Ducts
Puberty
Neo- Menarche Pregnancy
natal
Buds
Lobes
Terminal Buds
NulliParous
Lactation Peri Menop Post
menop
Menop
SPSS – LexiMine and Clementine
Puberty:
•Two hormones – estrogen and progesterone signal the
development of the glandular breast tissue.
•In female estrogen acts on mesenchymal cells to stimulate further
development.
•The gland increases in size due to deposition of interlobular fat.
•The ducts extend and branch into the expanding stroma.
•The epithelial cell proliferation and basement membrane
remodeling is controlled by interactions between the
epithelium and the intra-lobular hormone sensitive zone of
fibroblasts.
•The smallest ducts, the intra-lobular ducts, end in the epithelial buds
which are the prospective secretory alveoli.
•Breast ducts begin to grow and this growth continues until
menstruation begins.
Production of: Stroma, mesenchymal cells, epithelial cells
Reality of Disease
DNA
RNA Amino Acids
Genes
Gene Ontology
Proteins
Enzymes Substrates Co-Factors
Pathways
Tissues
Cells
Organelles
Processes: Tissue generation; Inflammation….
Physiological Systems
Physiological Development
(time)
4. Stratifying Disease
Tumor Staging
 T,M,N tumor scoring
 Analysis of Outcomes

Cancer Progression
localized
0
I
regional
metastatic
IIA IIB IIIA IIIB
IV
Tumor Progression
IIIA
IIA
0
I
IV
IIB
IIIB
Tumor Staging
Stage 0
(Tis, N0, M0)
Stage I
(T1,* N0, M0) ; [*T1 includes T1mic]
Stage IIA
(T0, N1, M0 ); (T1,* N1,** M0); (T2, N0, M0) [*T1 includes T1mic ]
[**The prognosis of patients with pN1a disease is similar to that of patients
with pN0 disease]
Stage IIB
(T2, N1, M0) ; (T3, N0, M0)
Stage IIIA
(T0, N2, M0); (T1,* N2, M0); (T2, N2, M0); (T3, N1, M0); (T3, N2, M0)
[*T1 includes T1mic ]
Stage IIIB
(T4, Any N, M0) ; (Any T, N3, M0)
Stage IV
(Any T, Any N, M1)
Stage IIIC
(Any T, N3, Any M)
10/10/02
T, M, N Scoring

T1: Tumor ≤2.0 cm in greatest
dimension
– T1mic: Microinvasion ≤0.1 cm in
greatest dimension
– T1a: Tumor >0.1 cm but ≤0.5 cm in
greatest dimension
– T1b: Tumor >0.5 cm but ≤1.0 cm in
greatest dimension
– T1c: Tumor >1.0 cm but ≤2.0 cm in
greatest dimension


T2: Tumor >2.0 cm but ≤5.0 cm in
greatest dimension
T3: Tumor >5.0 cm in greatest
dimension

N0: No regional lymph node
metastasis

N1: Metastasis to movable ipsilateral
axillary lymph node(s)

N2: Metastasis to ipsilateral axillary
lymph node(s) fixed or matted, or in
clinically apparent ipsilateral internal
mammary nodes in the absence of
clinically evident lymph node
metastasis
(T, M, N) Information Content
T
IIa
M
N
5. Tumor Heterogeneity
Breast tumors are heterogeneous
 Diagnosis primarily driven from H&E
 Co-occurrences of breast disease?
 Co-morbidities with other diseases?

2 3 5 7 8 1 1 2 2 2 2 2 2 2 2 3 3 3 3 3 3 4 4 4 4 4 4 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6 6
0 2 2 3 4 5 6 7 8 9 0 4 6 7 8 9 1 5 6 7 8 9 0 1 2 3 4 5 9 0 1 2 3 4 5 6 7 8 9
Bayesian Network of Diagnoses
Clinical Breast Care Project
Department of Defense
 20 % active duty personnel are female
 95 % active duty males are married
 Tri-Care health system

Clinical Breast Care Project

Collaboration between WRI and WRAMC
–
–
–
–
–
–
–

10,000 breast disease patients/year
Ethnic diversity; “transient
Equal access to health care for breast disease
All acquired under SINGLE PROTOCOL
All reviewed by a SINGLE PATHOLOGIST
2 military, 1 non-military site added 2003
6 military sites to be added 2006
Breast cancer vaccine program (her2/neu)
CBCP Repository
– Tissue, serum, lymph nodes (>15,000
samples)
– Patient annotation (500+data fields)
– Patient Diagnosis = {130 sub-diagnoses}
– Mammograms, 4d-ultrasound, PET/CT, 3T
MRI
– Complementary genomics and proteomics,
IHC
Current CBCP Studies






LOH vs tumor location
Modifier gene analysis in BRCA1/2
BC presentation in African Americans
Longitudinal Impact of Environmental/Lifestyle
MMG vs non-MMG detected BC and survival
Lymphedema
– Quantitative diagnosis (3d-ultrasound)
– Genomic and proteomic “risk” analysis

Mammography (GE, ICAD/CADx, SMDC)
– Breast density factors
– Integration of mammography and 3D ultrasound (“fusion”)
Studying Environmental Factors
Patients from JMBCC
In CBCP vs (CBCP-JMBCC)
CBCP
JMBCC
1.Scranton
2.Landstuhl
3.Japan
Windber Research Institute





Founded in 2001, 501( c) (3) corporation
Genomic, proteomic and informatics
collaboration with WRAMC
45 scientists (8 biomedical informaticians)
36,000 sq ft facility under construction
Focus on Women’s Health, Cardiovascular
Disease, Processes of Aging
WRI’s Mission
WRI intends to be a catalyst in the creation
of the “next-generation” of medicine,
integrating basic and clinical research
with an emphasis on improving patient
care and the quality of life for the patient
and their family.
WRI’s Core Technologies:
Central Dogma of Molecular Biology:
DNA  RNA  Protein
•
•
•
•
•
•
•
•
•
•
•
•
•
Tissue Banking
Histopathology
Immunohistochemistry
Laser Capture Micro-dissection
DNA Sequencing
Genotyping
Gene Expression
Array CGH
Proteomic Separation
Mass Spectrometry
Tissue Culture
Biomedical Informatics
Data Integration and Modeling
WRI Research Strategy
Cardiovascular Disease
Synergies
Obesity
CADRE
Lymphedema
CBCP
GDP
Women’s Health
Menopause
Aging
(2005)
WRI Partnerships
Amersham
Thermo-Finnegan
Waters
Teradata
MSA
Dept of Defense
USASMDC
Cimarron
InforSense
Oracle
MDR
GE Healthcare
ICAD/CADx
Correlogic
CiraSciences
Genomics/Proteomics
University of Pittsburgh
Infrastructure
Clinical Studies
Walter Reed Army Medical Center
University of Pittsburgh(UPMC, UPCI)
Georgetown University
Creighton University
University of Hawaii
Penn State University
Uniformed Services University-Health Sciences
UCSF- Breast Center
Preventative Medicine Research Institute
Pittsburgh Tissue Engineering Institute
University of Nevada-Las Vegas
Reasoning Environment
BioSim
BioWeb
BioSoft
Core 1 Computational Research
Core 2 Biomedical Research
Pedigree
Analysis
Patient
Synchron.
Core 3 Breast/Melanoma Risk
(Wen-Jen Hwu)
Driving
Biomedical
Projects
Disease
Stratific.
Information
Content
CoMorbidities
Race/Ethnicity
(Yudell)
Core 4 Infrastructure
Core 5 Training
Core 6 Dissemination
Pathway
Simulation
Data
Mining
Text
Mining
Co-Morbidity/Risk
(Esserman)
Data Integration

Data Warehouse Model
– Teradata  Oracle

Cimarron’s Scierra LIMS
– Amersham LWS
Creation of CLWS
 InforSense and SPSS

A Patient is:
Family History…… Nurse
Genomics…………. Genetic Couns.
Demographics…….Epidemiologist
Environment………Envir. Scientist
Patient
Lifestyle……………Social Scientist
Clinical History….. Physician
Therapeutic History.. Pharmacist
Tissue Samples……Pathologist
Cost of Treatment…Insurer
Quality of Life…….Patient
……….
A Patient is a Mother, Sister, Wife, Daughter…..
Modular Data Model






Socio-demographics(SD)
Reproductive History(RH)
Family History (FH)
Lifestyle/exposures (LE)
Clinical history (CH)
Pathology report (P)






Tissue/sample repository (T/S)
Outcomes (O)
Genomics (G)
Biomarkers (B)
Co-morbidities (C)
Proteomics (Pr)
Swappable based on Disease
Windber Storage Area Network
Hospital/WRI
Digital
Mammo
4d UltraSound
Pet/CT
3T MRI
PACS 1
PACS 2
PACS 3
PACS 4
Megabace
?
?
WRI
Hospital
MALDI
Windber SAN
?
NAS
CodeLink
Pathol
?
OC-3, OC-48
Pittsburgh
Philadelphia
Washington, DC
CLWS
WRI 7/2005
Conclusions



Personalized Disease will improve Patient Care,
Today; Personalized Medicine, Tomorrow
Disease is a Process, not a State
Translational Medicine must be both:
– Bedside-to-bench, and
– Bench-to-bedside

The processes of aging are critical:
– For accurate diagnosis of the patient
– For recognizing breast cancer as a chronic disease
Acknowledgements










Windber Research Institute
Joyce Murtha Breast Care Center
Walter Reed Army Medical Center
Immunology Research Center
Malcolm Grow Medical Center
Landstuhl Medical Center
Henry Jackson Foundation
USUHS
MRMC-TATRC
Military Cancer Institute
Patients, Personnel and Family!
m.liebman@wriwindber.org