In silico Evolution of Biological Clocks with Genetic Regulatory Networks

Genetic Regulatory Networks (GRNs) are the control systems of all cells. Their dynamics are of crucial importance in development [4] but also in the ongoing, reactive, metabolism [1]. A characteristic example of such responsive regulation are circadian rhythms, which were present already in early life forms [10]. Following Winfree [10, 11], we ask 1) How is it that biological clocks can adapt, within limits, to perturbations in cycle length, phase shift, and resetting? 2) Why in isolation do they run at internalized rates somewhat different from that of the external cycles? 3) Are these accidents of neutral selective value, or do they have some adaptive significance at the individual level? Evolving artificial genetic regulatory networks (aGRNs) that act as model biological clocks is a natural method to explore these questions. In our model, every network consists of a number of genes, each having any number of regulatory sites. Gene expression levels are determined by the activation of the corresponding sites and their interaction rules as well as gene type. Abstracting from transduction, environmental input simply raises the level of one protein type while the concentration of another type is read as output (fig. 1A, 2A). Starting from simple random networks, we use an evolutionary algorithm to arrive at a