Mutual Information and Information Flow in Biochemical Reaction Networks

Molecular abundances in cells vary due to the stochastic nature of individual biochemical reactions in cells. The resulting probability distributions are experimentally accessible and can be used to determine the mutual information between cellular components. However, even in simple biochemical pathways in which components form a causally ordered cascade, this mutual information is not necessarily monotonic. The possible increase of mutual information along signaling cascades can be intuitively understood in terms of noise propagation and time-averaging intrinsic fluctuations that are short-lived compared to an extrinsic signal. Mutual information measurements of stationary state distributions of molecular abundances thus seem of limited utility for characterizing cellular signaling processes. In contrast, we show how decomposing information flow in feedback loops can be used to rigorously establish general noise trade-offs in biochemical control networks.