Biology Reference
In-Depth Information
Additional pseudo-reactions can be specified as well.
These include non-reversible demand reactions that remove
a metabolite from the network. These are used for
compounds that are known to be produced for processes
beyond the scope of the model. In like manner, reversible
sink reactions can be added for metabolites that are
produced by reactions outside the scope of the recon-
struction but are necessary for metabolic network function.
Great care is taken with the addition of sink reactions, since
their addition may allow the model to grow without using
the metabolites in the medium. A visualization of these
reactions is shown in Figure 12.3 .
Once a network is converted into a mathematical model,
the model can be used to simulate how different pathways
are used under specific conditions. Classically, kinetic
information has been used to model small-scale metabolic
networks for which kinetic parameters are available.
However, such information is lacking for most enzymes.
Thus, large metabolic network models require simplifying
assumptions and alternative approaches to be amenable to
simulation. Flux-balance analysis (FBA) is one such method
for computational modeling of these networks [37] . This
technique assumes that the reaction rates (fluxes) within
a cell can reach a steady state represented mathematically as
S$v ΒΌ
consumption rates, product secretion rates, measured
internal flux constraints, etc.). In FBA, linear programming
is used to identify the fluxes for each reaction by optimizing
some objective function (e.g., biomass precursor produc-
tion). Figure 12.4 shows the procedure in more detail.
Stage 4: Network Evaluation
Once a reconstruction is converted to a model, the model is
analyzed and used for simulation. Simulation results are
validated against experimentally measured phenotypes.
This is done to provide confidence that subsequent model
predictions are accurate. Avariety of tests should be used to
increase model quality. Following each test, corrections to
the reconstruction are made as needed through iterative
rounds of manual curation, conversion to a model, and
network evaluation. This process continues until the
reconstruction is 'finished,' as determined by the ability of
the model to carry out the purpose and scope desired for
subsequent studies.
Topological Tests
The topology and stoichiometry of a metabolic network
clearly affect model functions. Thus, the first validation
done on a model involves verifying the mass balance of all
reactions within the metabolic network (except for
exchange, sink, demand, or biomass reactions). This is often
followed by an assessment of metabolic dead ends, which
are metabolites that are only produced or consumed. The
identification of these can help identify gaps, which are
model reactions that may be missing. Primary literature and
genome annotation are used to identify potential genes and
reactions present in the organism that would bridge identi-
fied gaps. Visualization of the metabolite's 'location' in the
metabolic network can be helpful for identifying potential
reactions that fill gaps in the network. Resources such as the
KEGG database [13] or biochemical textbooks/maps can
contain this information. Computational algorithms can aid
in this process by suggesting missing reactions, which must
be followed by experimental validation [38 e 40] . Care
should be taken to ensure that gaps are not inappropriately
0, where
S
is the stoichiometric matrix and
v
is the flux
vector in which each reaction j has a flux value
. In this
steady state, the requirements of mass-balance in chemical
reactions, combined with constraints on specific fluxes (e.g.,
flux of input exchange reactions) allow for the prediction of
reaction fluxes without requiring kinetic information.
Metabolic flux is often difficult to measure. This is
particularly true for non-central metabolic pathways. Thus,
the ability to predict flux for all pathways is particularly
valuable. The S matrix for a genome-scale model is under-
determined (i.e., fewer metabolites than reactions). This
means that genome-scale models will not provide a unique
solution specifying the flux of each reaction in the network.
However, they instead provide a range of feasible solutions,
often referred to as the 'solution space'. This solution space
contains all feasible steady-state flux values for each
reaction, subject to the metabolic network stoichiometry
and any user-inputted constraints (e.g., cellular metabolite
v j
FIGURE 12.3 A simplified representation of the various
reaction types present in metabolic models. Exchange reactions
allow the flow of metabolites into or out of the medium, transport
reactions allow the flow of metabolites into and out of the cell (or
between intracellular compartments). Demand and sink reactions
are for metabolites that are consumed or produced, respectively, by
reactions outside the scope of the metabolic model.
Exchange
Biomass
Sink
Other
Cellular
Processes
Metabolic
Model
Demand
Transport
 
Search WWH ::




Custom Search