Transcript Lecture_12_Drug_developmentx - Home | CISB-ECN

BMED 3510
Systems Biology in
Medicine and Drug Development
Book Chapter 13
Buzzwords: Personalized Medicine
And Predictive Health
“Personalized”:
Different treatments for different people.
Make custom-tailored predictions of one’s health “trajectory”
Sounds good, but are these goals feasible?
2
How Different Are We?
Genetic similarities
Naïve view: ~50% of genes from mom, ~50% from dad
Hence: brothers and sisters differ by ~ 50% (or not?)
Yet:
Genomes of humans and chimpanzees are very similar (98.8%!)
Two people should exhibit much greater similarity than that!
False deduction due to similarity between mom’s and dad’s genome
Differences mostly due to SNPs (single nucleotide polymorphisms) and
some genomic rearrangements.
If every 100th nucleotide could have a SNP and we have 3 billion
nucleotides, then we could have 430,000,000 different people!
3
Differences and their Importance
SNPs
Some important, others not really; often combinations are important
Other sources of variability
Some gene rearrangements, deletions, duplications
Epigenetics (DNA unchanged, but transcription (frequency) affected)
Important consequence for medicine
Every person potentially responds differently to treatment
Diseases and their treatments should be individualized, but are usually not
4
Personalized Medicine
Status quo:
Medicine is based on averages
(either from epidemiology or from animal experiments; later)
Task:
Need to progress from average input-output
correlations to a deeper understanding of
disease processes in individuals
Challenges:
1. Get the right data from individuals
2. Analyze them appropriately
(i.e., with (sophisticated) modeling)
Hope: Analogy with engineering
We do not need to take apart every machine
we encounter, if we understand the principles
that make this type of machine functional.
5
Personalized Disease Modeling
Concept
Develop dynamical model of a healthy system (person; pathway; …)
Determine parameter values
These are usually based on population averages
Replace average parameter values with person-specific values,
as much as possible
Study effects of the personalized combination of parameter values
Many parameter changes incur no significant symptoms
Scan the model for options of counteracting the disease
6
Example
Illustration Pathway
What is the same within a population; what is different?
Topology probably the same (What does it represent?)
Parameter values probably different (What do they represent? What if pi = 0)
How big are typical changes?
Do changes in parameter values make a big difference?
sensitivity analysis (typically one change at a time;
guaranteed for infinitesimally small changes)
simulations (simultaneous changes)
Homeostasis / allostasis
Personalized model
7
Example
8
Example
Simulation: @ t = 10, Vmax5 increased by 20%
5
x1
x2
x3
x4
x5
x6
2 .5
0
0
25
50
9
Example
Simulation: Start at steady state; @ t = 10, KI6 increased by 20%
5
x1
x2
x3
x4
x5
x6
2 .5
0
0
25
50
10
Example
Simulation: @ t = 10, h352 increased in magnitude by 20%
5
x1
x2
x3
x4
x5
x6
2 .5
0
0
25
50
11
Example
Simulation: @ t = 10, b6 increased in magnitude by 20%
5
x1
x2
x3
x4
x5
x6
2 .5
0
0
25
50
12
Example
Simulation: @ t = 10, KI6 and b6 increased in magnitude by 20%
5
x1
x2
x3
x4
x5
x6
2 .5
0
0
25
50
13
Example
Simulation: @ t = 10, h351 and h352 doubled
Run longer
x1
10
x2
x4
x5
x3
15
x6
5
7 .5
0
0
0
25
50
0
x1
x2
x3
x4
x5
x6
500
1000
14
Personalized, Predictive Health
Two issues
1. Identify differences between personal “parameters” and what’s “normal”
2. Investigate which differences (or combinations) are significant
3. Ideally identify significant difference before disease manifests
Search for biomarkers
Proteins (e.g., cytochrome p450 enzymes), genes, metabolites,
blood pressure, abnormal fingernails (kidney, liver, thyroid disease, …)
Big Q: Which biomarkers are symptomatic and which are causative?
15
Combinations of Biomarkers
One Biomarker:
(A to T) - SNP in HgbS
Sickle Cell Anemia
Many Biomarkers:
Oncotype DX Test (21 genes)
Remission of Breast Cancer
Hierarchical Networks of Biomarkers:
Disease
16
Biomarkers, Health and Disease Simplexes
One dimension: “normal range” (“U-box”)
normal
biomarker
Two dimensions: combined normal ranges
normal
biomarker 2
normal
biomarker 1
17
Biomarkers, Health and Disease Simplexes
Two dimensions: combined normal ranges + constraints
(Two extremes are not tolerable;
compensation between variables)
Result: linear bounds (reasonable approximation)
normal
biomarker 2
normal
biomarker 1
18
Biomarkers, Health and Disease Simplexes
normal
biomarker 2
normal
biomarker 2
normal
biomarker 1
normal
biomarker 1
19
Biomarkers, Health and Disease Simplexes
Many dimensions: polygon becomes a simplex
Note:
In principle, simplex can be computed from a model
20
Classification of Health & Disease
Ideal Solution (in full “biomarker space”):
Clear separation between health and disease simplexes
z
“Health
Simplex”
y
“Disease
Simplex”
x
21
Classification of Health & Disease
Would like to say: x <  : healthy; x >  : sick
(like PSA > 4)
z
y

x
22
Classification of Health & Disease
In reality, there is no unique  because disease status
also depends on other biomarkers, such as y and z.
z
y
“Healthy”
“Don’t know”

“Diseased”
Consequence: Looking at one biomarker insufficient
23
Health and Disease Trajectories (2-d)
Health
Temporary
Illness (fever,
dehydration, …)
Premorbidity
Treatable or
Self-healing
Disease
24
Health and Disease Trajectories
Health
Temporary
Illness (fever,
dehydration, …)
Premorbidity
Treatable or
Self-healing
Disease
25
Epidemiology
Molecular Biology
Biochemistry
Physiology
Hypothesized
Risk-Factor~Disease
Associations
Physiological
Mechanism
Process
Parameters
Personalized
Disease Models
Experimental
Systems Biology
Model Design
“Averaged”
Model
Perturbation
Numerical
Solution
Clinical
Trials
Simulation
Sensitivity,
Robustness
Personalized
Simulation
Health-Disease
Classification
Personalized
Risk Profile
Personalized
Health Model
Personalized
Health Prediction
Personalized
Treatment
Suggested
Prevention
“Averaged”
Treatment
Voit & Brigham, Open Path. J., 2008
Computational Systems Biology
26
Modeling in Drug Development
“Drug Development Pipeline”
Preclinical
Development
Discovery
Lead
Optimization
Target ID
Hit ID,
Lead ID
Note:
Clinical
Development
Clinical
Phase I
Development
of Drug
Candidate
Postclinical
Development
Clinical
Phase III
Clinical
Phase II
Launch
FDA
Approval
Process
1 NCE out of ~10,000 makes it; 10-20 years; ~ 1 Billion $
27
Modeling in Drug Development
Discovery
TI
Hit
Preclinical
Development
Lead
DC
Clinical
Development
CP1
CP2
CP3
Postclinical
Development
FDA
L!!
7. Launch
6. Seek FDA Approval
5. Test on large patient cohort
4. Test efficacy on small patient cohort
3. Test safety on healthy individuals
2. Optimize the most promising hit;
Formulate as drug with desirable properties
1. Identify biological target and molecules (potential drugs;
“hits”) that affect the target; screen for the most promising hit
28
Modeling in Drug Development
Discovery
TI
Hit
Preclinical
Development
Lead
relatively cheap
DC
Clinical
Development
CP1
CP2
CP3
Postclinical
Development
FDA
L!!
very expensive
Generic strategy: try to weed out as many molecules as possible
as early as possible, if they are not likely to make it to the end
Use models (and experiments) for screening process
29
Models in Drug Development
TI
Hit
Lead
DC
CP1
CP2
CP3
FDA
L!!
NCE (New chemical entity) screening
QSAR (Quantitative Structure-Activity Relationships)
Binding prediction (molecular dynamics)
30
Modeling of Receptor Binding
Many drugs work by binding to proteins. Here, the FDA approved drug Indinavir docks into the cavity
of an HIV protease in a lock-and-key mechanism (PDB: 10DW and 1HSG). Courtesy of Juan Cui and
Ying Xu, University of Georgia.
Modeling of Receptor Binding
TI
Hit
Lead
DC
CP1
?
Receptor
CP2
CP3
FDA
L!!
?
Antibody
Ligand
Modeling of Receptor Binding
100
L0=100
L [%]
C3
k+
k–
kR+
0
L
kL–
1
k2+
C2
0
k–
2
L0=1
time [days]
20
L0=100
k1+
R
10
100
kR–
C1
C3
kA–
L0=0.01
L
k3–
A
50
k+
3
Inject
L0=10
50
L0=10
L0=0.01
L0=1
0
0
10
time [days]
20
33
Compartment Models in Drug Discovery
TI
Hit
Lead
DC
CP1
CP2
CP3
FDA
L!!
k0B
kLB
kL0
B
L
IN
kBL
kB0
B  k0 B  kLB L  (kB 0  kBL )  B  IN
L  k B  (k  k )  L
BL
L0
LB
34
Compartment Models in Drug Discovery
k0B
kLB
kL0
B
L
IN
kBL
kB0
B  k0 B  kLB L  (kB 0  kBL )  B  IN
L  k B  (k  k )  L
BL
L0
LB
essentially linear;
easy to estimate from data
Models in Drug Discovery: PBPK
TI
Hit
Lead
DC
CP1
CP2
CP3
FDA
L!!
Blood
Much used model in pharmaceutical research:
Brain
Lung
Physiologically-Based Pharmacokinetic Model
First goal: Determine “ADME”: Absorption,
Distribution, Metabolism, and Excretion;
Fat
Extrapolation to other species
Kidney
Routes of drug administration
Liver
Dosage
36
Models in Drug Discovery: PBPK
Blood
Brain
Lung
Each organ and blood modeled as a
compartment with its specific features: Initially
a simple mass action model with influx,
retention, efflux.
E.g., fat tends to retain lipophilic drugs for a
longer time than lung. Volumes are taken into
account.
Fat
Kidney
Liver
Liver tends to degrade drug metabolically:
May include a metabolic model for this
compartment; may account for break-down
products.
Kidney, liver, lung provide possible exit routes
37
Pathway Screening
TI
Hit
Lead
DC
CP1
CP2
CP3
FDA
Concept:
Develop dynamic model of a physiological system
Introduce changes leading to disease
Systematically scan the model for means of disease treatment
L!!
Pathway Screening
TI
Hit
Lead
DC
Example: Simplified model
of purine metabolism
CP1
CP2
IG
GI
XU
H
U
GX
Uout
v In  5
X
L!!
H  v PH  v IH  v PHI  v HX  v Hout
X  vGX  v HX  v XU
U  v  v
I
G
FDA
P  v In  v PG  v PI  v PHI  v PH
I  v PI  v PHI  vGI  v IG  v IH
G  v  v  v  v
PG
P
CP3
v PG  320  P1.2  G 1.2
vGX  0.01  G 0.5
v PI  0.5  P 2  I 0.6
v IH  0.1  I 0.8
v PH  0.3  P 0.5
v HX  0.75  H 0.6
v PHI  1.2  P  I 0.5 H 0.5
v Hout  0.004  H
v IG  2  I 0.2  G 0.2
v XU  1.4  X 0.5
vGI  12  G 0.7  I 1.2
vUout  0.031  U
More Detailed
Purine Metabolism
R5P
vprpps
vpyr
vade
PRPP
Ade
Pi
vgprt
vpolyam
vden
vhprt
SAM
vaprt
vmat
vgmps
XMP
GMP
vtrans
vasuc
vimpd
IMP
vgmpr
S-AMP
Pi
Ado
AMP
ADP
ATP
vasli
vampd
GDP
GTP
vrnag
vgdrnr
Pi
vrnaa
RNA
vgrna
What does it take to set up
such a model?
varna
vadrnr
Lots of time and effort!!
Here:
vgnuc
vgprt
dGMP
dGDP
dGTP
vdnag
vdnaa
DNA
vgdna
vadna
vdgnuc
Pi
dAdo
dAMP
dADP
dATP
vhprt
vgua
HX
Ino
dIno
vhxd
vx
Over 30 variables
Dozens of parameters, …
vdada
vinuc
Gua
Guo
dGuo
vada
Xa
vxd
UA
vua
vhx
40
Curto et al., Math. Biosc., 1998
More Detailed
Purine Metabolism
R5P
vprpps
vpyr
vade
PRPP
Ade
Pi
vgprt
vpolyam
vden
vhprt
SAM
vaprt
vmat
vgmps
XMP
GMP
vtrans
vasuc
vimpd
IMP
vgmpr
S-AMP
Pi
Ado
AMP
ADP
ATP
vasli
vampd
GDP
GTP
vrnag
vgdrnr
Pi
vrnaa
RNA
vgrna
varna
vadrnr
vgnuc
vgprt
dGMP
dGDP
dGTP
vdnag
vdnaa
DNA
vgdna
vadna
vdgnuc
Pi
dAdo
dAMP
dADP
dATP
Typical analyses
HX
Ino
dIno
vhxd
vx
Xa
vxd
UA
vua
Do responses make sense?
Stability
Sensitivities
vada
vhprt
vgua
Diagnostics
vdada
vinuc
Gua
Guo
dGuo
Tasks:
vhx
Bolus experiments
Changes in enzyme activities
Changes in parameter values
Diseases
Treatments
41
Curto et al., Math. Biosc., 1998
More Detailed
Purine Metabolism
R5P
vprpps
vpyr
vade
PRPP
Ade
Pi
vgprt
vpolyam
vden
vhprt
SAM
vaprt
Suppose too much UA
vmat
vgmps
XMP
GMP
vtrans
vasuc
vimpd
IMP
vgmpr
Ado
AMP
ADP
ATP
vasli
vampd
GDP
1.
S-AMP
Pi
GTP
vrnag
vgdrnr
Pi
vrnaa
RNA
vgrna
varna
Explain:
e.g., PRPPS superactivity
or, HGPRT deficiency
vadrnr
vgnuc
vgprt
dGMP
dGDP
dGTP
vdnag
vdnaa
DNA
vgdna
vadna
vdgnuc
Pi
dAdo
dAMP
dADP
dATP
Intervene:
reduce UA production
vhprt
vgua
HX
Ino
dIno
vhxd
vx
2.
vdada
vinuc
Gua
Guo
dGuo
vada
Xa
vxd
UA
vua
vhx
3.
Side effects?
e.g.: UA   Xa 
42
Curto et al., Math. Biosc., 1998
Potential Application: Disease Simulators
Analogy: Flight simulator
Disease Simulator: Enter virtual person, symptoms, vital signs, disease
history, biomarker signals, … interactively simulate effects of drugs,
treatment options, …
43
Summary
Modeling has been used in drug development for some while
New approaches possible, due to better data and methods
Personalized medicine one of the hallmark goals of systems biology
Personalizing a model easy in principle (difficult in actuality)
Future: Disease simulators
44