Biology Reference
In-Depth Information
altering the architecture of the network itself under different
conditions or biological states, such as tissue type, disease
state or the surrounding environment
[118]
.
this network? Does this network support long-range inter-
actions, or is the impact of nodes contained locally? Are the
concentrations of the nodes in this network stable or gov-
erned by fluctuations? Will a small perturbation cause
a macroscopic failure? At least some of these characteris-
tics, and others like them, might be determined by the
topology of the network, regardless of the other details of
the interactions. And if that is the case, we should be able to
address these important questions with the tools of network
analysis.
In a broader perspective, applying graph theory to the
study of complex systems is aimed at bringing about an
intuitive, visual and mathematical toolkit for their under-
standing. In that sense, the challenge of this approach is to
devise a set of intuitive dynamical interpretations to the
already defined set of topological features. The idea is to
assign a functional meaning to characterizations such as
a broad degree distribution, high clustering, small world-
ness, etc. Along this path, future research needs to chal-
lenge some of the common wisdoms regarding these
attributions between structure and dynamics. For instance,
the intuitive notion that in a small world topology all the
nodes are affected by one another, since they are just a few
reactions away; or the common perception that the hubs are
the most influential nodes in the network. Once these
statements, and others like them, are examined, they will
bring forth a new intuition on the meaning of different
structural attributes. Then, by analyzing the structure of
a network, researchers will be able to make general
assessments regarding its expected dynamics.
The rapidly improving experimental techniques in
biology will hopefully enable the dynamical predictions
derived from network analysis to be tested. However, even
where the existing experimental procedures are insuffi-
cient, help might arrive from unexpected sources. Perhaps
the greatest success of the network approach thus far is in
revealing the universal nature of the topology of networks
e
cellular and others
e
providing a set of tools and criteria by
which to classify and characterize the structure of these
diverse systems
[5]
. A similar degree of universality in the
dynamics of networks, if found, will provide us with
a parallel set of unifying principles, allowing us to describe,
using a common platform, various dynamical processes and
make meaningful predictions on the behavior of networks
from diverse fields. These universal dynamical aspects
could then be inferred from one system to the other.
Metaphorically, this expands the boundaries of the classic
biology laboratories far beyond their traditional walls. As
data are currently collected in vast amounts from biolog-
ical, social and technological systems, the abilities that
network science creates to learn from one system about
the other provide a crucial source of empirical strength,
a strength that may one day help make complex systems
slightly more simple.
FROM STRUCTURE TO DYNAMICS
The path taken throughout this chapter outlines, in some
sense, the approach of network biology towards its future
challenges. We begin by describing the network represen-
tation of cellular systems, examining their topological
properties. We then follow with a discussion regarding
motifs, weighted networks and controllability, which
addresses the dynamics and function of these networks. In
this spirit, we end this chapter with what is probably the
most pressing challenge of this area of research
e
the
bridging between structure and dynamics. We are currently
at a stage where the topological aspects of cellular networks
have been thoroughly elucidated and their evolutionary
origins fairly understood. However, we still lack a complete
theory which could interpret the topological findings into
a set of dynamical predictions, from which the actual
functionality of the networks could be inferred
[5]
. Below
we stress, in a very broad fashion, the strategic path that
could meet this challenge
[119
e
120]
.
The most fundamental question we must address is
whether the gap between structure and dynamical behavior
could at all be bridged. We need to take into account that
the topology is one actor in a highly detailed cast of
network characteristics. In the most detailed description of
these cellular systems, all the interactions can take on
different reaction processes and different strengths. By
reaction processes we refer to the types of interaction, e.g.,
chemical, regulatory, etc., and by strengths we mean that
similar processes may occur at different rates. So whereas
structurally we denote all the various interactions by
network links, one should ask: is the process of genetic
regulation really comparable to that of physical binding? Is
it guaranteed that two structurally identical networks will
express similar behavior even if they differ in some other
details? Or perhaps these details, which are overlooked by
graph theoretic analysis, are not important.
A more constructive approach towards the above
questions is to ask how far can one actually progress with
structure alone? It seems clear that a complete time-
dependent dynamics of the system would require the
incorporation of all of the details mentioned above, and is
thus beyond the scope of network biology. On the other
hand, what could be achievable based on a structural
analysis is a macroscopic understanding of network
dynamics. More specifically, network science is not
expected to be successful in predicting the behavior of
a specific set of nodes. It could, however, provide answers
to general questions regarding the network as a whole.
Questions such as: which are the most effective nodes in
