Biology Reference
In-Depth Information
mediating signaling/regulation and transport events in the
cell. Genetic interaction (GI) networks deal with pairs of
proteins for which there is information that they interact
functionally (i.e., the absence or presence of both proteins
has a synergetic effect on the cell physiology/phenotype).
The BioGRID Database (
www.thebiogrid.org/
;
[144]
)is
the main repository of PPIs and GIs from a diverse selec-
tion of organisms, containing more than 60 000 unique PPIs
and more than 120 000 unique GIs for S. cerevisiae,
compiled and curated from more than 10 000 publications
(as of October 2011), with a fivefold increase in number of
PPIs, and a 15-fold increase in GIs since 2006
[145]
.
Topological investigations of the increasing number of PPIs
and GIs are ongoing
[145,146]
, together with studies
providing new insights into their architecture, interrela-
tionships and dynamics
[131,147
(GIM)
[167]
have been compared and combined, with
a consequent increase in the coverage and quality of GI
maps
[168]
.
Although not exempt from limitations, given that many
other interactions are occurring in the cell (e.g., see
Figure 18.1
), a comprehensive 'compendium of interac-
tions' with high predictive power was constructed by
integrating heterogeneous data with PPI and GI networks in
yeast. Park and co-workers integrated data from 30
different sources (or 'interaction' types) which included
functional similarity, co-expression, synthetic lethality
(GIs), PPI, phosphorylation and others
[169]
. In their study,
14 of the 20 novel genetic interactions proposed by their
prediction algorithm could be verified experimentally.
Studies of yeast interaction networks converted into
Boolean networks have also shown that phenotypes can be
predicted using the integrated interaction network RefRec
(
http://function.princeton.edu/bioweaver/
;
[170]
).
e
151]
. Clustering and
machine learning algorithms dominate the computational
methods developed for inference and analysis of PPIs and
GIs
[152]
, including hierarchical clustering, Bayesian
networks, and decision trees.
Correlation studies between transcriptional (RNA) gene
expression and PPI networks has revealed that co-expressed
genes are more likely to encode proteins that interact
[153]
,
with combined scoring of co-expression and PPI networks
revealing the order of interactions in a MAPK signaling
pathway
[154]
. More importantly, the topology and modu-
larity of yeast PPI networks ('interactome'), with prediction
of localization of proteins, and their relations with other
interactome networks in other organisms, from yeast to
higher eukaryotes and humans, with an essentially
conserved 'core' or essential proteins interactions,
complexes and modules conserved in all eukaryotes, and
their direct implications in molecular studies of human
diseases, is a field of continuous investigation and consti-
tutes one of the main challenges in the new era of systems
biology (
[7,16,17,45,102,155
e
Metabolic networks: Genome-wide reconstructed
models
Metabolic networks describe the relationships between
small biomolecules (metabolites) and the enzymes
(proteins) that interact with them to catalyze a biochemical
reaction. Metabolic networks, metabolic control and
modeling of metabolic networks in genome-wide recon-
structed models is a central area in systems biology
[16,17,52,55,171,172]
.
Non-linear and dynamic models are often used to
simulate metabolic and regulatory networks. However,
such models usually focus on one or few pathway(s) of
metabolism owing to technical limitations: e.g., the lack of
empirical information on kinetic parameters and/or reac-
tion mechanisms, high computational cost of non-linear
parameter estimation, and lack of dynamic experimental
data for verification of simulated results (for reviews
of available methods see
[121,173
157]
). To address this chal-
lenge, different groups integrating PPIs datasets with addi-
tional levels of biological information are advancing the
functional annotation of yeast genes
[158,159]
, the identi-
fication of new functional modules in the PPI networks
[160]
, and the capture of their dynamics during realistic
physiological perturbations, namely, the dynamics of stress
responses in yeast
[161]
. At the regulatory level, again,
phosphorylation/de-phosphorylation of proteins are major
regulatory events in the eukaryotic cell, and the yeast
phospho-proteome (PhosphoGRID;
www.phosphogrid.org/
;
[162]
), i.e., a network of proteins that interact with protein
kinases and phosphatases, was integrated with global PPI
networks
[163,164]
, revealing novel regulatory modules and
patterns of interactions.
At the genetic interactions (GI) networks level, datasets
obtained from three different experimental methods
e
176]
). Stoichiometric
models derived from metabolic networks are often genome
wide and hence can be readily integrated with high-
throughput data, Here, we will focus mainly on simulations
with linear and steady-state stoichiometric models, with
examples from yeast primary metabolism (i.e., catabolism
and anabolism, with secondary metabolism not covered;
[55]
).
Metabolic networks essentially map the proteins/
enzymes (often edges of the networks) interacting with
metabolites (often nodes of the network), to catalyze
metabolic reactions or to transport metabolites. Construc-
tion of such networks starts with the availability of the
genome sequence of the organism (i.e., S. cerevisiae;
www.
yeastgenome.org
; which provides the list of annotated
enzymes) which is progressively refined with the aid of
literature information
[17,171,177,178]
. A mathematical
e
e
synthetic genetic array (SGA)
[165]
, epistatic miniarray
profiling (E-MAP)
[166]
, and genetic interaction mapping
