Shaping the Information Channel: Molecules, Cells, and Experiments

Abstract

An important and ubiquitous question in the physics of biological systems is how an information transmission channel is shaped and optimized by a careful exploitation of the structural details of the underlying system. This dissertation explores three instances of this central question, at the levels of molecules, cells, and scientific experiments.

The first chapter is focused on the molecular gateway of cellular signaling, where a ligand concentration reflects some extracellular condition, and a receptor works as an information channel by binding the ligand and consequently activating the downstream signaling pathway. In this sense, the ligand-receptor binding event is the initial information channel of the signaling process, setting an upper bound on the final amount of information transmitted. We investigate how the information flow through the ligand-receptor binding can be optimized by the choice of the kinetic parameters, and how it is limited by various constraints of the cellular environment.

Once the signal is initiated, it needs to be relayed until it reaches the final target in the cell. Such signal transduction pathways are built on a network of specific protein-protein interactions, and one of the important determinants of interaction specificity is shape complementarity. In the second chapter, we aim to characterize the statistical properties of the ensemble of proteins in the cell, in terms of the shapes of protein surfaces. We study the intrinsic dimensionality of the space of surfaces, and discuss how it is linked to the properties of individual protein surfaces, revealing the non-trivial organization of the shape space.

The third chapter concerns the optimization of the design of a scientific experiment, viewing the experiment as an information channel through which the scientist collects data about the natural world. Specifically, we consider a behavioral neuroscience experiment where the aim is to infer the psychometric function, which governs the stimulus-dependent decision-making behavior of an animal. We demonstrate how the experimental design can be optimized to reach the desired precision of measurement with a minimal amount of data, using an adaptive, closed-loop algorithm that selects the most informative stimulus at each trial.