| Biochemical signal transduction, broadly defined as the conversion of the concentration of an input signal to an output response by networks of interacting proteins, is the most basic level of biological information processing. The biochemical reactions comprising these networks are inherently stochastic, inevitably effecting the faithful transmission of information. A major challenge is posed by the fact that input signals often possess dynamical and widely varying properties, and signaling networks are required to detect and process information under a broad range of environmental conditions. This thesis addresses how reliably signals can be transmitted through networks in the presence of noise.;The chemotactic response of the single-celled organism Escherichia coli (E. coli) has emerged as a prototype for understanding the physical principles underlying key properties of biochemical signaling networks, such as adaptation to the constant level of an input stimulus, robustness of the output to noise in network parameters, and the physical limits of measuring concentrations of signaling molecules. The network and its response are amenable to controlled and quantitative biochemical and genetic assays, resulting in detailed and experimentally faithful biochemical models. As such, chemotaxis provides an ideal experimental and theoretical model system for understanding not just how biochemical networks work but why they work the way they do.;In this thesis, we use a realistic, stochastic model to investigate information processing by the E. coli chemotaxis network. We compute for the first time the dominant kernels, or filters, employed by the network in processing the input. We show that the input-output relation of the network changes with, or adapts to, the statistical properties of the input signal. We demonstrate that the rescaling property of the observed input-output relation maintains the information about the stimulus transmitted by the chemotaxis network. |
