Biology Reference
In-Depth Information
sites is via chromatin immunoprecipitation, followed by
hybridization to a microarray (ChIP-chip)
[160]
or
sequencing (ChIP-seq)
[161]
. Targeted binding information
for TFs can also be found using a yeast-1-hybrid system see
Chapter 4). However, these approaches do not provide
information about whether the TF activates or represses
transcription of its target. Thus, gene expression analysis of
TF KO strains is needed. Owing to the combinatorial
interactions between TFs that act on some ORFs, gene
expression assays in multiple-TF KO strains
[162]
or ChIP-
seq assays for a TF in a different TF KO strain are needed to
fully elucidate TF connectivity. If this information is
available for major TFs in an organism under various
conditions, a putative TRN can be constructed in a fairly
automated fashion
[3]
. This is not to say that homology-
based genomic data
[163]
or databases
[164]
cannot be
useful in providing information for the reconstruction.
However, experimental data are an important source of
information and a method for algorithmic generation of
such data has been described
[165]
.
Although the generation of TRNs is better understood,
it is more difficult to convert them into computable models
that accurately recapitulate transcriptional regulation.
However, a variety of methods are being attempted
[166
e
169]
. Generally, TRNs are represented as a network,
with TF nodes connected to their ORF targets
(
Figure 12.6
A). For small-scale networks, stochastic
[170]
and kinetic
[171]
models are readily computable, but they
do not scale well for genome-scale models
[3]
.
TRNs have been integrated with metabolism through
a few approaches. In one approach called regulatory FBA
(rFBA) the TRN rules were represented as a Boolean
network that activated or repressed the flux through meta-
bolic reactions associated with the genes the TFs regulated.
This was applied to a simplified model of core metabolism
and used for simulation
[172]
. Since the Boolean network
needs to converge to a steady state, the simulations were
conducted through an iterative series of calculations with
the TRN and metabolic network. Each iteration started by
first imposing the Boolean rules based on metabolite
concentrations in the growth medium. FBA was then used
to optimize growth with the metabolic network, and
subsequent changes in extracellular metabolite concentra-
tion were computed based on the uptake and secretion
profiles from the model. These concentrations were then
used to set regulatory states of genes for the next iteration.
Interestingly, this approach was successful in modeling
phenomena controlled by regulatory mechanisms such as
diauxies
[50]
and was later expanded to use ordinary
differential equation-based modeling of TRN dynamics
[173]
. Current efforts to represent genome-scale TRNs in
a computable fashion rely largely on Boolean network
representation
[167]
approaches using concepts such as log-linear kinetics
[3,174
e
176]
. Integration with other network types using
these representations (for non-genome-scale TRNs)
[162,168,173,177]
has been explored and was found to
improve the predictive accuracy of the models.
Non-system-level understanding of regulatory networks
has been used to guide metabolic engineering of yeast for
increased galactose consumption
[178]
. Thus, a systems-
based approach for metabolic engineering via TRN modi-
fications may prove beneficial. Other applications include
the use of integrated models to identify potentially novel
regulatory interactions to explain discrepancies between
experimental and predicted data, as was accomplished for
Halobacterium salinarum NRC-1
[179]
, E. coli
[162]
, and
S. cerevisiae
[177]
.
Toward a Whole Cell Model
A major goal in systems biology is the attainment of a whole
cell model that incorporates the function of every gene for an
organism. Such a model might be amenable to interrogation
of all biological processes of the organism to understand how
all components are coordinated to yield the cell phenotype.
Integration of the metabolic network (
Figure 12.6
B) with the
O and E networks discussed above would theoretically result
in the OME network (
Figure 12.6
D). However, the best
mathematical representation of such a network on a genome-
scale is an open question. Beyond this integration, the simu-
lation of all processes within a cell would also require the
inclusion of signaling networks, sRNA regulation
[180]
,
protein modifications, and account for cell division mecha-
nisms. Ultimately, understanding the reactions behind each of
these networks will allow representation in a stoichiometric
matrix similar to those for the M and E matrices, making
integration an easier task.
GLOSSARY
Biomass reaction A pseudo-reaction consisting of the summation
of all biomass precursors
in their appropriate fractional
distribution.
Demand reaction Reactions added to the model to facilitate
consumption of a metabolite when the actual consuming reaction
is unknown or outside the scope of the model.
Exchange reaction Reactions added to the model
to supply or
remove metabolites from the in silico 'medium'.
Extreme pathway The boundary/edge vectors of the solution space
to
v¼
0. Each is linearly independent of the others and any point
in the solution space can be described as a linear combination of
extreme pathways.
Flux balance analysis A method of analysis for finding an optimal
solution to
S
S
v¼
0 using linear programming. This method relies
on optimization for a user-specified reaction, such as the biomass
reaction.
and additive kinetic modeling
