Transcript No Slide Title

Computational Modelling
of Biological Pathways
Kumar Selvarajoo
[email protected]
Outline
• Background of Research
• Methodology
• Discovery of Cell-type Specific Pathways
• Analysis of Complex Metabolic Diseases
The levels in Biology
DNA
Organism
transcription
The
Central
Dogma of
Molecular
Biology
RNA
Organ
translation
Protein
Tissue
Cell
Is Genome Sequence Enough?
•
The genome sequence contains the information for living systems
propagation
•
The functioning of living system involves many complex molecular
interactions within the cell
•
How do we understand these complex interactions with static sequence
information?
From Genome to Cellular Phenotype
Successful
Sequence Analysis
Genome
Sequence
Eg. Human
Functional
Mapping
Gene/Protein
Function
Eg. ESR
Coding
????
Cellular
Networks
Eg. Glycolysis
Tissue
Phenotype
Eg. Cancer,
Diabetes
The steps involved to convert genome sequence
into useful phenotypic description
From Genome to Cellular Phenotype
•
Understanding the individual function of genes, proteins or metabolites
does not allow us to understand biological systems behaviour
•
It is therefore important to know how each gene, protein or metabolite
is connected to each other and how they are regulated over time
•
Recent technological breakthroughs in biology has made generating
high throughput experimental data a reality
•
But by analysing high throughput experimental data of biological
systems without understanding the underlying mechanism or circuitry
is not very useful
Computation in Biology
•
Computational methods hence become essential to help understand
the complexity of biological systems (Hartwell et al, Nature,1999)
•
However, the currently available computational techniques are
insufficient to accurately model complex biological networks (Baily,
Nature Biotechnology, 2001)
•
This is mainly due to the general lack of formalised theory in biology at
present.
•
Biology is yet to see its Newton or Kepler (Baily, Nature Biotechnology,
2001)
Advantages: Computer Simulations
•
Easy to mathematically conceptualise
•
Able to develop and predict highly complex processes
•
Rapid creation and testing of new hypotheses
•
Serves to guide wet-bench experimentation
•
Potential cost reductions with accelerated research
Simulation Techniques
• ‘Bottom-Up’
–
–
–
–
–
Predominant in biology (e.g. Enzyme Kinetics)
Deliberately COMPREHENSIVE (include everything)
Need lots of experimentally determined parameters
Very long process
Very expensive
• ‘Top-Down’ or ‘Phenomic’
– Common in engineering
– Deliberate use of APPROXIMATIONS (reduce complexity)
successful in engineering (e.g. Finite Element Analysis)
– Very fast
– Inexpensive
Problems with ‘Bottom-Up’
Approaches
Metabolic Network
• The correlation between mRNA levels and
protein expression levels are very poor
• Protein post-translational modifications cannot
be predicted from the genome sequence
Proteins
• The kinetic parameters used to determine the
rate of protein activity is very difficult to determine
mRNA
• In vitro determination of kinetic parameters fail
to capture the robustness of biological systems
found in vivo
Genomic
Sequence
• Even if all parameters are determined, the
model is not versatile or scalable, that is, usually
only applied to one cell-type at one specific
condition (e.g. muscle cells at aerobic condition)
‘Top-Down’ Approach
Metabolic Network
Proteins
• Attempt to develop a network module*, hence
cannot be comprehensive
• First look at a well known network and try to
understand the topology through phenotypic
observation
• Formulate the interactions within the network
with guessing parameters for protein activity
mRNA
• Check with experiments once parameters are
fixed
• Perform perturbation experiments to confirm the
hypothesis
Genomic
Sequence
• Useful for drug perturbation studies
*A functional module is, by definition, a discrete entity whose function is separable from those of other modules.
(Hartwell et al, 1999, Nature)
Modules
Metabolic
Networks
in
We chose the
glycolytic module
Our Methodology
Knowing the true system
k
A
B
C
Systems Approach
SB  fn( A, C, X )
X
Our Methodology
Consider a simple (ideal) reaction, one mole of substrate A converted to
one mole of product B by the enzyme E1
E1
A
Assume
B
S A  e kt
S B 1  e
 kt
Our Methodology
In a typical enzymatic reaction (non ideal), physical constraints exist
that prevent complete depletion of substrate. Therefore,
S B  kf (1  e
 kbt
)
where kf is the fitting parameter and 0< kf<1 (Constraint)
Our Methodology
For feedback/feedforward mechanisms k2 could be a function of the
upstream/downstream substrate
k2
A
B
X
SB  k1 sin t  k 2(Sx)(1  ek 3t )ek 4t  k 5SA
Constraints
•
Constraints are introduced to increase the coefficient confidence
•
Examples
- lead coefficient
- rate coefficient
- frequency coefficient
Constraints
Lead coefficient constraint, 0< kf<1
E1
A
B
S B  kf (1  e
 kbt
)
Constraints
Rate coefficient constraint, 0.1<kb<1.0
1.2
1
Concentration (mM)
0.8
kb=1.0(max)
0.6
kb=0.1(min)
increasing
kb
0.4
0.2
0
0
1
2
3
4
5
Tim e (s)
6
7
8
9
10
Features of Our Methodology
• Fewer parameters required
• Able to construct complex networks
• Able to produce accurate predictions even under
reduced complexity
• Uses and predicts metabolite concentrations, rather
than enzyme activity
Glycolytic Network and Measured Values for
Erythrocytes (RBC)
Comparison between Measured
and Predicted Values in RBC
A
B
A/B
Metabolites
(-)
G6P
0.039
0.0390
1.00
F6P
0.013
0.0129
1.01
FBP
0.0027
0.0027
1.00
DHAP
0.14
0.1400
1.00
G3P
0.0057
0.0058
0.98
BPG
0.0007
0.0007
1.00
3PG
0.069
0.0705
0.98
2PG
0.01
0.0106
0.94
PEP
0.017
0.0180
0.94
PYR
0.085
0.0881
0.96
*Model of 2,3-biphosphoglycerate metabolism in the human erythrocyte
Biochem. J. 342 (1999), Mulquiney & Kuchel
*
Measured* (mM) Predicted (mM)
Robustness of Model Parameters
+/- 20% Variation in Input G6P Values
0.18
0.16
0.14
Concentration (mM)
0.12
0.1
sim
-20%
20%
0.08
0.06
0.04
0.02
0
G6P
F6P
FBP
DHAP
G3P
BPG
Glycolytic Metabolites
3PG
2PG
PEP
PYR
Robustness of Model Parameters
+/- 20% Variation in All Model Parameters
0.16
0.14
0.12
Concentration (mM)
0.1
sim
-20%
20%
0.08
0.06
0.04
0.02
0
G6P
F6P
FBP
DHAP
G3P
BPG
Glycolytic Metabolites
3PG
2PG
PEP
PYR
Model Application
• Model applied to other cell types and conditions
• These are predictions - No experimental data from
the ‘test’ cell type is used (unless stated otherwise)
• Model parameters are fixed unless stated otherwise
• Points of accurate prediction represented by green,
otherwise indicated as red
Metabolic Phenotypes of Erythrocytes
and Myocytes are Highly Distinct
3
2.5
2
1.5
1
ln(Ratios)
0.5
0
G6P
-0.5
-1
F6P
FBP
G3P
3PG
2PG
PEP
PYR
Prediction of Myocyte Glycolytic
Phenotype
0.5
G6P(IM)
F6P
FBP
DHAP
G3P
BPG
3PG
2PG
PEP
PYR
0.45
Model: Reference (RBC)
Test: Myocytes
IM:
0.45mM
0.4
Concentration (mM)
0.35
* data not available
0.3
sim
0.25
exp
0.2
0.15
0.1
0.05
*
*
0
G6P
F6P
FBP
DHAP
G3P
BPG
Glycolytic Metabolites
3PG
2PG
PEP
PYR
Discovery of Cell-type Specific Pathways
Using Computational Simulations
Trypanosoma Brucei (T.brucei)
• is a parasite
• causes the African Sleeping
Disease or Trypanosomiasis
• carried by Tsetse fly
Prediction of T.brucei Glycolytic
Phenotype (Aerobic Condition)
25
G6P(IM)
F6P
FBP
DHAP
G3P
BPG
3PG
Concentration (mM)
20
2PG
PEP
PYR
Model: Reference (RBC)
Test: Tbrucei
IM:
1.64mM
15
sim
exp
10
5
0
G6P
F6P
FBP
DHAP
G3P
BPG
Glycolytic Metabolites
3PG
2PG
PEP
PYR
Prediction of T.brucei Glycolytic
Phenotype under Aerobic Condition
25
G6P(IM)
F6P
FBP
DHAP
G3P
BPG
2PG
PEP
PYR
Model: Reference w ith
m odfication at DHAP, G3P & BPG
Test: Tbrucei
IM:
1.64m M
20
Concentration (mM)
3PG
15
sim
exp
10
5
0
G6P
F6P
FBP
DHAP
G3P
BPG
Glycolytic Metabolites
3PG
2PG
PEP
PYR
Comparison of Predicted T.brucei Glycolytic
Phenotype Against a Literature Model*
30
FBP
25
Concentration (mM)
20
sim
exp
15
lit
10
5
0
G6P
F6P
FBP
DHAP
G3P
BPG
3PG
2PG
Glycolytic Metabolites
*Glycolysis in Bloodstream Form Trypansoma brucei
J. Bio. Chem, 342 (1997), Bakker B. M. et al
PEP
PYR
Optimising model for Cell-Specificity,
T.brucei
Prediction of T.brucei Glycolytic Phenotype after
Optimisation, Aerobic Condition
25
G6P(IM)
F6P
FBP
DHAP
G3P
BPG
Concentration (mM)
20
3PG
2PG
PEP
PYR
GLY3
Model: Reference
(Tbrucei, Aerobic)
Test: Tbrucei
(Aerobic)
15
sim
exp
10
5
0
G6P
F6P
FBP
DHAP
G3P
BPG
3PG
Glycolytic Metabolites
2PG
PEP
PYR
GLY3P
Prediction of T.brucei Glycolytic Phenotype under
Anaerobic Condition
25
G6P(IM)
F6P
FBP
DHA
BPG
G3P
3PG
Concentration (mM)
20
2PG
PEP
PYR
GLY3
Model: Reference
(Tbrucei, Aerobic)
Test: Tbrucei
(Anaerobic)
15
sim
* data not available
exp
10
5
*
*
0
G6P
F6P
FBP
DHAP
G3P
BPG
3PG
Glycolytic Metabolites
2PG
PEP
PYR
GLY3P
Aerobic Condition
T.brucei