Computer simulated evolution of a network of cell-signaling molecules.

We have trained a computer model of a simple cell-signaling pathway to give specified responses to a pulse of an extracellular ligand. The pathway consists of two initially identical membrane receptors, each of which relays the concentration of the ligand to the level of phosphorylation of an intracellular molecule. Application of random "mutational" changes to the rate constants of the pathway, followed by selection in favor of certain outputs, generates a variety of wave forms and dose-response curves. The phenotypic effect of mutations and the frequency of selection both affect the efficiency with which the pathway achieves its target. When the pathway is trained to give a maximal response at a specific concentration of the stimulating ligand, it gives a consistent pattern of changes in which the two receptors diverge, producing a high-affinity form with excitatory output and a low-affinity form with inhibitory output. We suggest that some high- and low-affinity forms of receptors found in present-day cells might have originated by a similar process.