Biology Reference
In-Depth Information
particularly informative and well-studied network topology
parameter is node connectivity, or 'degree'. Protein
may result from genetically induced perturbation of
protein interactions
pro-
tein interactome networks follow a power-law distribution,
where most nodes have a low connectivity and a few nodes,
the network 'hubs', have a high connectivity (see Chapter
9). This 'scale-free' degree distribution is also observed in
many other networks, the internet being a particularly
noteworthy example, and has important consequences for
network robustness. The overall structure of a scale-free
network is hardly affected by the removal of random nodes,
but is highly sensitive to removal of hubs
[175]
. This scale-
free behavior has clear applications in the design of
electrical power grids, but what does
183]
. Similarly,
when a single gene is linked to multiple disorders, it often
seems to be because distinct mutations of this gene affect
specific individual interactions with different partners
[94,184]
. It follows that looking for 'disease interactions'
rather than for 'disease genes' should assist the delineation
of disease mechanisms and aid efforts to rationally inter-
rupt disease progression.
From an interactome network perspective the effects
of genetic variations are traditionally modeled as
complete loss of gene products ('node removal'). While
such interpretations are generally suitable for nonsense
or frameshift mutations occurring early in the protein,
large insertions or deletions, or complete gene knock-
outs, the node removal model may not readily apply to
truncations that preserve specific autonomous protein
domains, or to single amino acid substitutions. Such
genetic variations could instead lead to perturbations of
specific
[59,166,178,181
e
e
it mean for
protein
protein interactome networks?
About a decade ago, it was reported that hubs in the
protein
e
protein interactome of S. cerevisiae are prefer-
entially essential, meaning that knockouts of the corre-
sponding genes tend to be unviable
[176]
. This initial
observation was biased by the inclusion of literature-
curated interactome maps, where 'star' proteins that have
an artificially high connectivity also tend to be essential
[39,176]
. There is instead unbiased evidence of a correla-
tion between degree and essentiality in co-complex
interactome networks, but not in binary interactome
networks
[39,177]
. Deeper examination clearly demon-
strated a correlation between connectivity in an inter-
actome network map and functional pleiotropy in
S. cerevisiae
[39]
. Thus, the connectivity of a protein does
relay information about the phenotype of its correspond-
ing gene. Together with other topological information,
network connectivity may one day be used to predict new
disease genes, as suggested by the observation that
proteins associated with cancer are preferentially hubs in
the human interactome
[178]
.
e
interactions
('edge
removal'), or
'edgetic
perturbations'
[184,185]
.
The systematic isolation of genetic variants associated
with edgetic perturbations, or edgetic alleles, and their
characterization in vivo, represent a promising strategy for
investigating the function of specific interactions, partic-
ularly with regard to human disease
[185]
. Two comple-
mentary strategies, 'forward and reverse edgetics',
reminiscent of the time-tested dichotomy of forward and
reverse genetics
[186]
, allow systematic investigation of
the phenotypic outcomes of perturbations of specific
binary protein interactions
[187]
.Takingasetofmuta-
tions in a gene associated with particular phenotypes,
such as disease-associated mutations, the forward edget-
ics approach uses Y2H to determine the interaction
defects of proteins where the mutations have been intro-
duced by site-directed mutagenesis
[184]
. Reciprocally,
reverse edgetics starts from a set of interactions for
a protein of interest, and aims to systematically isolate
alleles encoding proteins with desired specific interaction
defects by reverse Y2H selections
[44,68,188]
.The
edgetic alleles that are thus selected can be reintroduced
in vivo into a model organism to investigate the pheno-
typic consequences of specifically altering the corre-
sponding molecular interaction(s)
[185]
.
Besides individual edges, higher-level topological
structures such as network motifs can also be associated
to specific biological functions
[135,136]
. Different types
of networks exhibit distinctive profiles of the relative
abundance of network motifs, so network motif profiles
can be used to characterize and compare networks.
Neuron networks or regulatory networks are enriched
in feedforward loops, whereas food webs are enriched
in bi-parallel motifs
[135,136,189]
. These distinctive
enrichments suggest
Assigning Functions to Individual
Interactions, Protein Complexes
and Network Motifs
Functional genomics experiments and function prediction
algorithms are typically designed to uncover the biological
roles of genes and gene products. We argue that if inter-
actome networks underlie genotype
phenotype relation-
ships, then edges (protein interactions) should be
associated with functions and phenotypes just as nodes
(proteins) are.
Fanconi anemia (FA) is a rare chromosome instability
disorder associated with congenital defects and suscepti-
bility to cancer. Of the 13 genes genetically associated
with FA, seven encode members of a core FA protein
complex
[179,180]
. This example and others show high
interconnectivity between proteins associated to a partic-
ular disease, which suggests that the disease phenotype
e
that
interactions that are part of
