Using random perturbations to infer the structure of feedback control in gene expression

Abstract

Feedback in cellular processes is typically inferred through cellular responses to experimental perturbations. Modular response analysis provides a theoretical framework for translating specific perturbations into feedback sensitivities between cellular modules. However, in large-scale drug perturbation studies the effect of any given drug may not be known and may not only affect one module at a time. Here, we analyze the response of gene expression models to random perturbations that affect multiple modules simultaneously. In the deterministic regime we analytically show how cellular responses to infinitesimal random perturbations can be used to infer the nature of feedback regulation in gene expression, as long as the effects of perturbations are statistically independent between modules. We numerically extend this deterministic analysis to the response of average abundances of stochastic gene expression models to finite perturbations. Across a large sample of stochastic models, the response of average abundances generally obeyed predicted bounds from the deterministic analysis, but dramatic deviations occurred in systems with bimodal or fat-tailed stationary state distributions. These discrepancies demonstrate how deterministic analyses can fail to capture the effect of perturbations on averages of stochastic cellular feedback systems--even in the linear response regime.