Biology Reference
In-Depth Information
diseases, as well as disease genes, tend to divide into
modules, or families, of related disorders. In the disease
gene network genes that contribute to the same disorder
tend to be correlated in many other ways as well. They have
an increased tendency to be expressed together in specific
tissues; they typically display high co-expression levels;
and in many cases they share common cellular and func-
tional characteristics, as annotated in the Gene Ontology
[86]
. This network is also in close relationship with the
protein
e
protein interaction network, as disease-related
genes are very likely to have their products interact together
through physical binding. A surprising feature revealed by
this analysis is the distinction between lethal genes and
disease genes. In many cases, the products of lethal genes
are highly connected nodes in the protein interaction
network
[41]
. This emphasizes the importance of the hubs
for the proper function of the network. In contrast, disease
genes tend to avoid the hubs, and the vast majority of them
are non-essential. It has been suggested that this is driven
by natural selection, enabling the proliferation of mutations
only if they harmed non-vital genes
[86]
. Confirming
evidence comes from the fact that somatic mutations,
which do not harm the organism's reproduction, are indeed
more frequently related to hub genes. In a broader
perspective, this network approach to human diseases
offers a tool for the understanding of general patterns in
genetic disorders, and could potentially reveal connections
which are not apparent in the study of individual disorders
[87
e
91]
.
responsible for the execution of some basic biological
operations. Topologically, as discussed earlier, this is
expressed in the emergence of various sub-graphs
composed of highly interlinked groups of nodes. The high
clustering typical of cellular networks provides the quan-
titative evidence for this modular network structure. In this
section we focus on the typical recurring structures of these
sub-graphs, and their meaning. In a sense, we are lowering
the altitude of the bird's-eye perspective with which we
viewed the networks until now, going from the macroscopic
analysis to a more focused look at the building blocks of
our complex systems.
Sub-graphs and Motifs
In order to conduct a fruitful analysis of network modules,
we need first to develop a scheme by which we can identify
what are meaningful modules. For instance, consider
a tetrahedral sub-graph, which is a fully connected set of
four nodes as shown in
Figure 9.4
(a). We can evaluate the
abundance of this sub-graph in our network, but this will
not be sufficient in order to tag it as a significant functional
module. The randomness in the network topology makes it
probable that such a module is due to appear in the network
by chance. We thus consider a certain module to be
a significant motif if it is over-represented in that network,
that is, more abundant than expected by chance alone
[92
e
93]
. The idea is that if the network has the tendency to
over-represent a certain module, there must be an evolu-
tionary or functional need for it. Since natural selection
discriminates on the basis of functional criteria, it will be
these motifs that are likely to be capable of carrying
important biological functions.
THE BUILDING BLOCKS OF CELLULAR
NETWORKS
In the previous section we discussed the macroscopic
aspects of the hierarchical topology. We have shown that
the hierarchy in cellular networks is closely intertwined
with their modular structure. Indeed, from a functional
point of view biology is full of examples of modularity,
where a distinct group of proteins, genes or metabolites is
Randomized Networks
As stated above, for a certain module to qualify as a motif it
must be more abundant than would be expected by chance.
However, we have not accurately defined what we mean by
(A)
(B)
(C)
(D)
X
X
Y
Y
X
Z
Z
FIGURE 9.4
Network motifs: (A) The hypothetical tetrahedral motif; (B) the autoregulator; (C) the coherent feedforward loop and (D) the incoherent
feedforward loop.
