Quantifying information transfer in biochemical networks

Abstract: Living cells possess remarkable information processing capabilities, ranging from cell-fate determination to chemotaxis. These capabilities arise from biochemical reaction networks, whose stochastic dynamics can be described mathematically using counting processes. Quantifying information transfer in such systems, however, poses both computational and conceptual challenges. Exact information rates are difficult to obtain even for simple networks, motivating Gaussian small-noise approximations that are expected to become accurate at large molecular copy numbers. Yet these approximations can yield qualitatively incorrect information rates even in this regime. In this talk, I will discuss how this discrepancy arises from the distinction between reaction- and concentration-based descriptions of biochemical trajectories. Conventional Gaussian descriptions combine the contributions of distinct reaction channels into effective concentration fluctuations, thereby discarding information carried by individual reaction events. I will use Duncan’s theorem to derive reaction-based Gaussian information rates that agree with those of the discrete counting-process model in the large-volume limit. I will then connect this time-domain formulation to the Fourier-space approach, which allows these rates to be calculated conveniently from the spectral densities of the input and output signals. Finally, I will discuss the conceptual and biological implications of this distinction: under what conditions can downstream biochemical networks access information carried by individual reaction events? I will show how active reaction cycles can make this information accessible through molecular concentrations and connect these results to the limits of cellular sensing.

MathematicsPhysics

Audience: researchers in the topic

Probability, Statistical Mechanics and Quantum Fields