Optimality in sensing and prediction

Abstract

Biological systems are optimized for particular functions, but still exhibit significant diversity along some axes. This thesis focuses on the intersection of optimization and variability, with a focus on biological functions related to sensing and processing information.Chapter 2 describes work performed with Stephanie Palmer, where we considered the prediction problem inherent in circadian clocks. Starting from a simple encoding framework, we are able to explain why circadian clocks often do not have 24-hour periods, and to make explicit experimental predictions about cyanobacterial circadian clocks. Chapter 3 is based on work in collaboration with Benjamin Hoshal, Kyle Bo- janek, Jared Salisbury, Olivier Marre, and Stephanie Palmer. In this work, we used electrophysiological recordings from a larval salamander retina to infer the under- lying population structure, and were able to show that this population structure is functionally valuable for identifying scenes. Chapter 4 describes work done with William Bialek. We asked if there is a function-level explanation for why some systems have highly precisely arranged pho- toreceptors, and others have more variable arrangements. We found that high fidelity signal transmission only imposes a very weak constraint on photoreceptor arrange- ment, and that other constraints are necessary to explain the highly precise cases. Chapter 5 is based on work with Stefan Landmann and Mikhail Tikhonov. We considered prediction problems that single-celled organisms might face, such as pre- dicting fluctuations in environmental resources in a structured environment. We showed that known circuit motifs can be minimally modulated to allow for this kind of prediction.