The functioning of cellular regulatory networks is jointly governed by deterministic nonlinear system dynamics and stochastic fluctuations. They can amplify or neutralize each other, and their effects are often difficult to separate, particularly, in the context of feedback regulation.
We show that besides cooperativity, even the ubiquitous protein dimerization can generate sufficient nonlinearity to support bistable cellular memory in positive feedback loops. We developed a method to split these feedback loops at the molecular level by minimizing signal distortion. The resulting open loop relations captured the nonlinearities and together with equivalence relations, they mapped the parameter space of deterministic bistability for the positive feedback circuits. Considerably larger was the measured parameter space of the corresponding stochastic activity, characterized by random transitions between the two states of the bimodal expression. Noise and deterministic transients predicted this expansion of the stochastic transitions and also their rates.
Since noise and deterministic responses have different origins, feedback splitting can help identify the molecular mechanisms that underlie the dynamics of dosage compensation, cellular differentiation and reversal of cell fates.