Systems Biology Models: Challenges and Applications

1
Systems Biology ModelsChallenges and
Applications

• Michael Breen

National Center for Computational Toxicology U.S.
Environmental Protection Agency Research Triangle
Park, North Carolina
SAMSI Transition Workshop, May 16, 2007
2
Outline

• Background
• Applications
• Approaches
• Challenges
• Example Model of ovarian steroidogenesis to
  predict biochemical response for baseline and
  fadrozole studies
• In vitro steroidogenesis assay with ovary
  explants
• Ovarian steroidogenesis model with enzyme
  inhibition by fadrozole
• Steady-state analysis
• Estimation of parameters
• Assessment of model fit
• Sensitivity analysis

3
Institute for Systems Biology

• Internationally renowned non-profit
  research institute founded in 2000
• 170 cross-disciplinary faculty and staff composed
  of biologists, engineers, mathematicians,
  statisticians, computer scientists, and
  physicists
• Develop new methods and technologies to enable
  systems approaches to disease
• Awarded 10M from Bill Melinda Gates Foundation
  in 2005

Institute for Systems Biology, Seattle, WA
4
Enabling (Omic) Technologies
Cell Biology
Transcription
Translation
Metabolism
mRNA
Protein
DNA
Metabolites
Mass Spectrometry
Protein gels
Microarrays
5
Features of Systems Biology

• Global measurements
• Measure changes in transcripts, proteins, etc.
  across state changes
• Create mathematical systems models that integrate
  different data types
• DNA, RNA, protein, interactions
• Measure dynamic changes across development,
  physiological disease, or environmental exposures

6
Outline

• Background
• Applications
• Approaches
• Challenges
• Example Model of ovarian steroidogenesis to
  predict biochemical response for baseline and
  fadrozole studies
• In vitro steroidogenesis assay with ovary
  explants
• Ovarian steroidogenesis model with enzyme
  inhibition by fadrozole
• Steady-state analysis
• Estimation of parameters
• Assessment of model fit
• Sensitivity analysis

7
P3 Medicine
Predictive, Preventative, and Personalized (P3)
Medicine

• Driven by systems approaches to disease and new
  measurement technologies (e.g. in vivo molecular
  imaging)
• Develop drugs to cure disease (reengineer
  disease-perturbed biological networks)
• Develop drugs to prevent disease (prevent
  biological networks from becoming
  disease-perturbed)

8
Dose-Response Modeling
9
Dose-Response Modeling
Mechanism of Action
Mechanistic Mathematical Models
10
Outline

• Background
• Applications
• Approaches
• Challenges
• Example Model of ovarian steroidogenesis to
  predict biochemical response for baseline and
  fadrozole studies
• In vitro steroidogenesis assay with ovary
  explants
• Ovarian steroidogenesis model with enzyme
  inhibition by fadrozole
• Steady-state analysis
• Estimation of parameters
• Assessment of model fit
• Sensitivity analysis

11
Approach

• Carefully choose tractable biological models
• Consider current technologies and complexity of
  biological system
• Focus of each system biological response to
  perturbations
• Perturbation genes respond physiological
  changes
• Modularization
• Iterative process

12
Modular description of a cell
http//www.gnsbiotech.com/biology.php
13
Outline

• Background
• Applications
• Approaches
• Challenges
• Example Model of ovarian steroidogenesis to
  predict biochemical response for baseline and
  fadrozole studies
• In vitro steroidogenesis assay with ovary
  explants
• Ovarian steroidogenesis model with enzyme
  inhibition by fadrozole
• Steady-state analysis
• Estimation of parameters
• Assessment of model fit
• Sensitivity analysis

14
Challenges

• Identification of network topology
• Parameter values
• Level of detail
• Lumping
• Relative versus absolute measurements
• Coordination!!!!

15
Coordination Collaboration

• Expose fish to chemicals
• Global measurements
• Transcriptomics microarrays, RT-PCR
• Proteomics 2D-gels, mass spectrometry
• Metabonomics NMR
• Integrate omics data to estimate model
  parameters

Small fish exposure system
Fathead minnows
16
Outline

• Background
• Applications
• Approaches
• Challenges
• Example Model of ovarian steroidogenesis to
  predict biochemical response for baseline and
  fadrozole studies
• In vitro steroidogenesis assay with ovary
  explants
• Ovarian steroidogenesis model with enzyme
  inhibition by fadrozole
• Steady-state analysis
• Estimation of parameters
• Assessment of model fit
• Sensitivity analysis

17
Effects of EAC on HPG Axis
Hypothalamus
GnRH
Anterior Pituitary
Negative Feedback
LH, FSH
Gonads (Ovaries, Testes)
T, E2
Androgen/Estrogen Responsive Tissues (e.g.
liver, gonads)
18
Effects on Steroid Metabolism
I
I
I
Steroidogenesis metabolic pathway
Fadrozole Inhibit CYP19 Breast
cancer therapy Trilostane
Inhibit 3ßHSD Cushings disease treatment
19
Mechanistic Computational Steroidogenesis Model

• Improve understanding of dose-response behavior
  for EAC
• Help define mechanism of actions for poorly
  characterized chemicals
• Serve as a basis to identify predictive
  biomarkers (patterns of steroid changes)
  indicative of adverse effects
• Support environmental human health and ecological
  risk assessments
• Help screen drug candidates based on steroid
  effect in early phase of drug development

Chemical Dose
Response
Chemical Dose
Mechanism of Action
Response
20
Population Effects Model
Coupled Systems Model
HPG-Axis Systems Model
Coupled Differential Equations
Population Model
Steroidogenesis Model
Statistical Model
Mechanistic Models
21
Endocrine Disruption in Fish

• Convincing evidence that fish are affected at
  individual and population levels
• Fish may serve as effective environmental
  sentinels for possible effects in other
  vertebrates
• Evolutionarily conserved HPG axis

Fathead minnow
22
Objective
Create a computational model of ovarian
steroidogenesis and estimate parameters to
predict synthesis and secretion of T and E2 for
in vitro baseline and fadrozole studies
23
Outline

• Background
• Applications
• Approaches
• Challenges
• Example Model of ovarian steroidogenesis to
  predict biochemical response for baseline and
  fadrozole studies
• In vitro steroidogenesis assay with ovary
  explants
• Ovarian steroidogenesis model with enzyme
  inhibition by fadrozole
• Steady-state analysis
• Estimation of parameters
• Assessment of model fit
• Sensitivity analysis

24
In Vitro Steroidogenesis Experiments Baseline

• Dissect fish ovary
• Incubate ovary in medium supplemented with
  cholesterol
• Collect medium at six time points over 31.5 hr
• Measure medium concentrations of testosterone (T)
  and estradiol (E2) using radioimmunoassay

Small fish culture facility
Fathead minnows
25
In Vitro Steroidogenesis Experiments Fadrozole

• Dissect fish ovary
• Incubate ovary in medium supplemented with
  cholesterol and five fadrozole (FAD)
  concentrations
• Collect medium at 14.5 hr
• Measure medium concentrations of testosterone (T)
  and estradiol (E2) using radioimmunoassay

Small fish culture facility
Fathead minnows
26
Outline

• Background
• Applications
• Approaches
• Challenges
• Example Model of ovarian steroidogenesis to
  predict biochemical response for baseline and
  fadrozole studies
• In vitro steroidogenesis assay with ovary
  explants
• Ovarian steroidogenesis model with enzyme
  inhibition by fadrozole
• Steady-state analysis
• Estimation of parameters
• Assessment of model fit
• Sensitivity analysis

27
Conceptual Steroidogenesis Model

• 6 unique enzymes
• 12 enzymatic reactions
• 4 secreted steroids

28
Computational Steroidogenesis Model

• 6 transport rates
• 12 first-order enzymatic reaction rates
• 2 enzyme inhibition constants

29
Dynamic Mass Balances
Ovary
Net metabolic rate
Net uptake rate

• Yields a system of coupled differential equations
• 20 model parameters

30
Competitive Inhibition
E S
E P
E S
31
Outline

• Background
• Applications
• Approaches
• Challenges
• Example Model of ovarian steroidogenesis to
  predict biochemical response for baseline and
  fadrozole studies
• In vitro steroidogenesis assay with ovary
  explants
• Ovarian steroidogenesis model with enzyme
  inhibition by fadrozole
• Steady-state analysis
• Estimation of parameters
• Assessment of model fit
• Sensitivity analysis

32
Measured Steroids from Baseline Study
R2 0.95
R2 0.98
R2 0.84
R2 0.94

• Good evidence steroid synthesis is operating near
  steady-state during experiments
• Steady-state assumption reduces model complexity

33
Steady-State Analysis

• Set differential equations to zero to yield
  algebraic equations
• Determined analytical solutions for testosterone
  (CT,med) and estradiol (CE2,med) using Maple
  software
• Solutions depend on 11 out of 20 parameters

where
34
Steady-State Analysis
(9)
35
Outline

• Background
• Applications
• Approaches
• Challenges
• Example Model of ovarian steroidogenesis to
  predict biochemical response for baseline and
  fadrozole studies
• In vitro steroidogenesis assay with ovary
  explants
• Ovarian steroidogenesis model with enzyme
  inhibition by fadrozole
• Steady-state analysis
• Estimation of parameters
• Assessment of model fit
• Sensitivity analysis

36
Parameter Estimation
Cost function
Model-Predicted
Model-Predicted
Measured
Measured
where
measured testosterone for d th FAD dose at i th
time model-predicted testosterone measured
estradiol for d th FAD dose at i th time
model-predicted estradiol measured fadrozole
for d th FAD dose

• Applied an iterative optimization algorithm
• Simultaneously estimated parameters using data
  from baseline and fadrozole-exposure studies

37
Estimated Parameters
Ovary Uptake of Cholesterol and Fadrozole
Secretion of Testosterone and Estradiol
k10 k18 k19
k0 k15
1726.553 149.301 102.171
hr-1 hr-1 hr-1
15401.470 0.0015
pg ml-1 hr-1 Partition coefficient
(dimensionless)
First-order Enzyme Kinetics with Inhibition by
Fadrozole
0.509 5.8 3.2
356.217 8143.017 4671.198
hr-1 hr-1 hr-1 hr-1 pg ml-1 pg ml-1
k9 k11 k12 k13 k16 k17
Literature values from fish experiments
FAD inhibition constants
Breen MS et al. Annals of Biomedical Engineering,
2007
38
Outline

• Background
• Applications
• Approaches
• Challenges
• Example Model of ovarian steroidogenesis to
  predict biochemical response for baseline and
  fadrozole studies
• In vitro steroidogenesis assay with ovary
  explants
• Ovarian steroidogenesis model with enzyme
  inhibition by fadrozole
• Steady-state analysis
• Estimation of parameters
• Assessment of model fit
• Sensitivity analysis

39
Evaluation of Model Fit Baseline Study
Breen MS et al. Annals of Biomedical Engineering,
2007
40
Computational Steroidogenesis Model

• Expect E2 to decrease with increasing FAD
• Expect T to increase with increasing FAD

41
Evaluation of Model Fit Fadrozole Study
Breen MS et al. Annals of Biomedical Engineering,
2007
42
Outline

• Background
• Applications
• Approaches
• Challenges
• Example Model of ovarian steroidogenesis to
  predict biochemical response for baseline and
  fadrozole studies
• In vitro steroidogenesis assay with ovary
  explants
• Ovarian steroidogenesis model with enzyme
  inhibition by fadrozole
• Steady-state analysis
• Estimation of parameters
• Assessment of model fit
• Sensitivity analysis

43
Sensitivity Analysis
Relative Sensitivities
where
model-predicted testosterone model-predicted
estradiol i th parameter

• Analytically determined partial derivatives with
  respect to each parameter
• Evaluated relative sensitivities for control and
  each fadrozole dose

44
Sensitivity Analysis
Testosterone
Estradiol
Breen MS et al., 2007

Dose-dependent sensitivity
45
Sensitivity Analysis
46
Summary

• Steroidogenesis model can predict T and E2
  concentrations, in vitro, while reducing model
  complexity with steady-state assumption
• Sensitivity analysis indicated E1 pathway as
  preferred pathway for E2 synthesis
• Mechanistic model could be useful for
  environmental risk assessments and drug
  development with chemicals that alter activity of
  steroidogenic enzymes

47
Dynamic Steroidogenesis Model

• 17 first-order enzymatic reaction rates
• 2 enzyme inhibition constants
• 14 reversible steroid transport rates

48
Acknowledgements
Cross-ORD Mentors Rory Conolly, NCCT Haluk
Ozkaynak, NERL Gerald Ankley, NHEERL
NC State University, Biomathematics Program,
Dept. of Statistics Miyuki Breen
Small fish steroidogenesis
EPA Duluth, MN Daniel Villeneuve Dalma
Martinovic Elizabeth Durhan Kathy Jensen Michael
Kahl Elizabeth Makynen
EPA Cincinnati, OH David Bencic Iris
Knoebl Mitchell Kostich James Lazorchak David
Lattier Gregory Toth Rong-Lin Wang
EPA STAR Program Karen Watanabe (Oregon Health
Sciences Univ.) Nancy Denslow (University of
Florida) Maria Sepulveda (Purdue
University) Edward Orlando (Florida Atlantic
University)
Pacific Northwest National Laboratory Ann Miracle
EPA Athens, GA Tim Collette Drew Ekman Quincy
Teng
US Army Vicksburg, MS Edward Perkins
EPA Grosse Isle, MI David Miller
H295R cells steroidogenesis
EPA, NHEERL RTP, NC Leonard Mole Sidney
Hunter Ralph Cooper John Laskey Jerome Goldman
Mitsubishi Pharma Corporation, Chiba,
Japan Natsuko Terasaki Makoto Yamazaki
University of Saskatchewan, Saskatoon,
Canada Markus Hecker John Giesy