Biology Reference
In-Depth Information
representation of metabolic networks can be used in
simulations to predict the fluxes through the reactions,
assuming that the culture is at a steady state (none of the
metabolites accumulate in the cells and the extracellular
growth medium is of a constant composition). The stoi-
chiometry of the reactions in the model is usually known,
hence these reactions can be listed as linear equations of
mass conservation. Using metabolic flux analysis (MFA; or
flux-balance analysis, FBA)
[179,180]
, these equations are
then used as constraints of a linear optimization problem,
with selected objectives such as maximization of biomass
production, minimization of oxygen or energy consump-
tion, or other targets of interest. Experimental data on
metabolite levels are also used as constraints to narrow
down the space of possible solutions (each one a set of
fluxes for all the reactions in the model). A variety of
constraint-based methods have been developed for simu-
lations with stoichiometric models
[179,181,182]
.
The genome-wide metabolic network reconstructions
have been distributed in non-standard formats until the
SBML
[183]
format became available. The BioModels
database
[184]
is one of the databases that store and
distribute the models. The tools available for simulation of
metabolic models are either toolboxes/packages developed
for use with general computational tools (e.g., SBMLTool-
box/COBRA/MATLAB;
[185,186]
), (SimBiology/MAT-
LAB) or standalone platforms such as BioOpt (see BioMet
online toolbox;
www.sysbio.se/BioMet
;
[187]
) and COPASI
(
www.copasi.org/
;
[188]
).
Simulation of yeast metabolism via constraint-based
methods has been integrated with information from
viability of deletion mutants
[189
generate additional knowledge (e.g., identification of genes
responsible for catalyzing specific reactions in the yeast
metabolic network;
[204,205]
).
Metabolic models of S. cerevisiae, as a model eukaryote,
can be extended with the integration of signaling/regulation
events
[102]
and dynamics
[201]
as suggested by Bailey
a decade ago
[206]
, providing an optimum platform for the
global investigation of metabolism
[52,55]
. Comparison and
integration of yeast metabolic and regulation networks with
PPI networks from an evolutionary point of view based on
their topological properties has revealed the importance of
temporal and spatial aspects
[200]
, and large metabolic
fluxes as a possible driving force
[207]
. Finally, the inte-
gration of kinetic data from quantitative proteomics and
metabolomics, the two levels most directly involved in
metabolic control, has been proposed as the most compre-
hensive way to integrate metabolomics in genome-wide
systems biology models
[55]
.
Network analysis/visualization tools
A search in the Nucleic Acids Research (NAR) directory of
biology-related web servers databases and tools (
http://
bioinfo
rmatics.ca/links_directory/
)
using the keyword
'interaction' brought
in 121 results (retrieved October
2011),
the majority of
these designed for
storing
protein
protein interaction (PPI) datasets, or predicting
novel PPIs, most of them with visualization tools. A
smaller subset of these has been designed to work with
arbitrary networks and is able to handle more than one type
of molecular interaction data.
Table 18.1
shows a non-
exhaustive list of the publicly available tools. As an
example, GEOMI (
[208]
;
http://www.systemsbiology.org.
au/downloads_geomi.html
) allows four-dimensional (4D;
i.e., three dimensions
e
191]
and metabolome
e
and transcriptome data (e.g.,
195]
). Constraint-
based methods have also been used to predict protein
localization
[196]
.
The topology of yeast metabolic networks has been
investigated using high-throughput data
[197,198]
, as well
as being compared and integrated with other networks
[102,164,199,200]
.
Simulations of metabolic models are usually based on
the assumption of steady-state cultures. On the other
hand, industrial yeast cultures are traditionally dynamic
batch or fed-batch cultures (e.g., wine production;
recombinant protein production processes). Apart from
other modeling strategies, the construction of a hybrid
model of yeast culture basedonthemetabolicnetwork
and a non-linear kinetic description of nutrient uptake
rates makes it possible to predict the behavior of dynamic
yeast cultures
[201]
.
Metabolic models are not limited to constraint-based
modeling
[202,203]
. Thus, metabolic modeling from graph
theory and logical modeling have been used as background
knowledge for hypothesis generation by robot scientists,
with experimentation and machine learning able to
[192
e
time-course dynamics) network
visualization of protein interactions, and has been applied
to the visualization and analysis of the S. cerevisiae com-
plexome network
[209]
. International community efforts
are progressively being developed to promote collabora-
tions and the implementation of new open source tools for
network-based visualization and networks analysis (e.g.,
Cambridge Networks Network, CNN,
http://www.cnn.
group.cam.ac.uk/
; National Resource
รพ
for Network
Biology,
http://www.nrnb.org
).
Towards Comprehensive Integration of 'Omics'
Datasets from Single Experiments
Data mining and processing of 'omics' datasets (e.g.,
transcriptome, proteome, metabolome) towards compre-
hensive integration (e.g., principal components analysis;
proteome
transcriptome correlations) should ideally be
performed with data extracted from the same experiment,
using standardized, curated sampling protocols. Also,
comprehensive integration will always require a prior
e
