314

23

Regulatory Networks

neer biomolecular assemblies for a variety of applications, including the creation of

polymeric materials. ⁷

23.2 Network Modelling

The proteins are considered as the nodes of a graph (cf. Sect. 12.2), and a pair

of proteins is joined by a vertex if the proteins associate with each other. The

“interactome”—the set of interactions in which a protein participates—is character-

ized by the graph (or an equivalent list of all the proteins in a cell, each associated

with a sublist of the proteins with which they interact). This is in contrast to metabolic

networks, in which two metabolites are joined if there is a chemical reaction (catal-

ysed by an enzyme) leading from one to another (Sect. 23.15). Attention is often

focused on small portions of these networks, which are then called pathways.

All proteins are, of course, gene products. ⁸ Hence, the fundamental regulatory

network is that of the genes, ⁹ which constitute the nodes, the edges signifying the

activation or inhibition of other genes, and the central problem is to infer (“reverse

engineer”) both the state structure of the network (cf. Fig. 12.1) and the physical

network of interactions. For the former, the input data are now typically the temporal

evolution of gene expression profiles, obtained by a succession of microarray exper-

iments (cf. Sect. 18.1). For the latter, association is measured more or less directly

using a variety of physicochemical techniques.

The graph of interactions is potentially extraordinarily large and complex (and,

let us reiterate, ignores higher-order correlations). Even if one confines oneself to the

upper NN expressed proteins in a cell, there aretilde upper N squared∼N ² potential binary interactions and vastly

more higher-order ones; ¹⁰ and even if only a small fraction of these interactions

actually occur (and some general results for the stability of systems (Sect. 12.1)

suggest that only about 10% will occur), we are still talking abouttilde 10 Superscript 7∼10⁷ interactions,

assuming about 10 Superscript 410⁴ expressed proteins (in a eukaryotic cell), and 10 Superscript 810⁸ pairs would

have to be screened in order to find the 10%. In a prokaryote, with possibly only

1000 expressed proteins, the situation is more tractable but still poses a daunting

experimental challenge, even without considering that many of those proteins are

present in extremely low concentrations.

When one or more stimuli arrive at a cell, the affinities of certain proteins for

a transcription factor binding site (TFBS) are altered, and mRNA transcription is

activated or inhibited, resulting in altered abundance of the mRNA and the trans-

7 Wilson et al. (2018).

8 This statement, the obvious corollary of the central dogma, is actually quite problematical—in

the sense of having a rather ambiguous meaning—when scrutinized in detail. Many functionally

relevant proteins are significantly modified (e.g., glycosylated) by enzymes after translation. Of

course, the enzymes themselves are gene products.

9 Schlitt and Brazma (2005).

10 Many transcription factors, for example, are multiprotein complexes.