Lecture 4 Dec 2000, Per Kraulis
Metabolic and signalling networks are complex: There are many nodes (proteins, particles, molecules) and many connections (interactions). This is obvious from nearly any illustration of a network. And even if one can define sub-networks that can be meaningfully described in relative isolation (such as the citric acid cycle, or the ras p21 signalling network), there are always connections from it to other networks.
The overall structure of a network can be described by several different parameters. For example, the average number of connections a node has in a network, or the probability that a node has a given number of connections. Theoretical work has shown that different models for how a network has formed will give different values for these parameters.
The classical random network theory by Erdös and Renyi (1960) states that given a set of nodes, the connections are made randomly between the nodes. This gives a network where most nodes have the same number of connections. Recent research has shown that this model does not fit the structure found in several important networks, such as the Internet, World-Wide Web, or social networks.
Instead, these complex networks are better described by a so-called scale-free, hierarchical model where most nodes have only a few connections, but a few nodes (called hubs) have a very large number of connections. Very recent work by Jeong, Barabasi et al (Nature (2000) 407, 651-654), indicate that metabolic networks are examples of such scale-free networks. This result is important, and will probably lead to new insights into the function of metabolic and signalling networks, and into the evolutionary history of the networks.
Robustness is another important property of metabolic and signalling networks. This is the ability of the network to produce essentially the same behaviour even when the various parameters controlling its components vary within considerable ranges. For instance, recent work (von Dassow et al, Nature (2000) 406, 188-192) indicates the segment polarity network in the Drosophila embryo can function satisfactorily with a surprisinly large number of randomly chosen parameter sets. The parameters do not have to be carefully tuned or optimized. This makes biological sense: A signalling system should be tolerant of mutations or environmental changes.
One important aspect of metabolic and signalling networks is that they are strongly conserved in evolution. Generally, the same roles in the metabolic networks in different organisms are performed by homologous enzymes, with the same kind of regulatory mechanisms. Of course, there are cases where a non-homologous protein is involved in one organism compared to the norm in others. There are also also cases where an entire section of a network is missing in an organism.