Exploiting fluctuations in gene expression to infer causal interactions between genes

Characterizing gene regulatory networks to gain a mechanistic description of cellular behaviours is a key challenge in systems biology. Identifying causal interactions between genes is traditionally done with perturbation experiments. However, precise perturbation experiments with known targets are not always possible while keeping cells in their physiologically relevant regime. We present a novel approach to use naturally occurring stochastic fluctuations in cells to infer causal interactions between genes, without the need to perturb cells or follow cells over time. Our approach exploits the fact that gene expression noise propagates through causally connected genes. This noise propagation can be identified through violations of a co-variability condition using a passive reporter. Our theoretical results promise to harness the information contained in cell heterogeneity data obtained from single-cell sequencing and flow cytometry experiments. We thus report experimental data to quantitatively test our theoretical results, using well-characterized synthetic gene regulatory circuits in E. coli.