The function of certain proteins is to interact with the "regulatory regions", generally found upstream from the genes, and to switch those genes on or off - in other words to regulate their expression. Directly or indirectly, the products of these genes are then likely to have an effect on the expression of other genes. This creates networks of molecular interactions which adapt protein production to the cell's needs in a given context. A knowledge of these networks is of great importance, because it could explain cell specialisation within a multicellular organism, and more generally, its development and morphogenesis.

At present, these interactions are described in the specialist literature, and less often in databases, in the form of text, diagrams or graphs. But it will only be possible to check how consistent they are, to compare them against gene expression data, or to simulate their function, if they exist in an computer-compatible form. Unfortunately, not enough is known at present about these interaction networks to allow them to be modelled in detail, for example in the form of systems of differential equations whose variables would be a function of the concentrations of different products. Only basic models are possible. In the simplest form, a network can be represented as a set of Boolean variables, with a value of 1 or 0 according to whether the corresponding gene is expressed or not, and a set of transitions between the values of these variables. Despite its simplicity, a model of this kind, which has more sophisticated variants, is capable of exploring the dynamics of interaction networks, for example to predict the existence of feedback circuits or steady states. This allows the analysis and simulation of well-defined systems such as the network of ten genes involved in flower formation in the plant A. thaliana  [1].

Other models are currently being developed, with the aim of producing more realistic behaviour patterns by including graduated information about reactions, of the type "the more gene A is expressed, the less gene B is expressed". At the same time, technological progress in "DNA chips" heralds the availability of vast amounts of gene expression data. Thanks to these chips, it will be possible to work out the structure of underlying interaction networks, although this will need the help of methods of analysis which have not yet been devised. Within the same timescale, there are already ambitious projects aiming to create models linking the genomic and metabolic levels  [2].