Stochastic bimodalities in deterministically monostable reversible chemical networks due to network topology reduction

Recently, stochastic simulations of networks of chemical reactions have shown distributions of steady states that are inconsistent with the steady state solutions of the corresponding deterministic ordinary differential equations. One such class of systems is comprised of networks that have irreversible reactions, and the origin of the anomalous behavior in these cases is understood to be due to the existence of absorbing states. More puzzling is the report of such anomalies in reaction networks without irreversible reactions. One such biologically important example is the futile cycle. Here we show that, in these systems, nonclassical behavior can originate from a stochastic elimination of all the molecules of a key species. This leads to a reduction in the topology of the network and the sampling of steady states corresponding to a truncated network. Surprisingly, we find that, in spite of the purely discrete character of the topology reduction mechanism revealed by "exact" numerical solutions of the master equations, this phenomenon is reproduced by the corresponding Fokker-Planck equations.