Biology Reference
In-Depth Information
5. Interpreting Network Topology
Clustering and motif analysis represent two important ways of making
use of protein networks, beyond the simple information integration and
browsing that networks offer generically. Another, much-studied aspect
of networks lies in their higher-level topology. How many interactions
exist in the network as a whole? How are these distributed on the nodes?
Are there universal structuring principles in biological networks? In a
series of very influential papers,
3,39-42
Albert-László Barabási and his col-
leagues have drawn attention to a number of unique topological features
of biological networks. These features have since been found in many
types of networks; in fact, they seem to be observable almost universally.
First, biological networks are almost never random networks. A ran-
dom network would be one where a certain number of edges (interac-
tions) have simply been randomly distributed among a number of nodes
(proteins). Such networks usually would have a characteristic average
“degree”, whereby degree is defined as the number of edges emanating
from a given node. For a random network, the average degree simply
depends on the number of edges assigned; and for a large fraction of the
nodes, the degree will correspond closely to the average degree of the
entire network. In contrast, real biological networks have a very different
distribution of degree values: most nodes have only a few edges (low
degree), and a small number of nodes have very many edges (high
degree). In fact, the distribution of degree values is often observed to
closely follow a power law, such that the number of nodes with degree
k
is inversely proportional to
k
taken to some characteristic power
d
(degree exponent). Such networks do not have a characteristic average
degree, at which most nodes would have their number of connections;
while an average degree can indeed be computed, it is rather meaning-
less. Therefore, such networks have been termed “scale-free”.
Interestingly, not only biological networks are found to be scale-free —
the same is true not only for many technical networks (e.g. the Internet,
airline networks), but even for social networks such as friendships or mat-
ing networks. In each of these networks, the highly connected nodes
serve as “hubs”, and only upon their removal does the network as a
whole suffer dramatic consequences. Most non-hubs can be removed
