Natural selection as the process of accumulating adaptive information.

Abstract

A defining feature of life is its ability to respond to complex conditions by leveraging information about the environment. In many respects, evolutionary biology is interested in understanding how genetic information changes over time to produce organisms that are exquisitely adapted to their environments. The core insight of the modern synthesis is that this adaptive information is acquired by the process of natural selection acting on genetic variation. While population genetics has developed a large body of theory regarding how genetic variance changes in the process of evolution, we lack correspondingly rich theory describing how adaptive information changes as a consequence of these same dynamics. In Chapter 1, I aim to clarify the sense in which the genome encodes adaptive information about particular environments, establish and interpret Shannon's mutual information as the appropriate measure of adaptive genetic information, and illustrate how different evolutionary forces shape this information. In addition, I demonstrate that this view of genetic information is not only of theoretical interest, but also practical and useful as a measure of adaptive differentiation. Results from classical population genetics suppose that there is a cost of selection, measured in terms of substitutional load, that limits how much information can be gained by this process. In Chapter 2 I test, extend, and reinterpret this theory to illuminate a rigorous and meaningful relationship between information gain and fitness under very general conditions. By reframing the process of natural selection in learning theoretic terms, I am able to clarify and formalize the cost of selection in terms of regret---a relative measure of load. I then establish general bounds on this cost and show that there is indeed a fundamental fitness cost associated with information acquisition by selection. These results highlight the centrality of information acquisition to natural selection and the value that information-theoretic perspectives have in evolutionary biology. Overall, this work contributes to an understanding of selection that advances evolutionary theory by synthesizing population genetics, information theory, and learning theory in modeling the acquisition of adaptive information. In the final year of my Ph.D., the world witnessed the outbreak of the COVID-19 pandemic. In response to this crisis, I shifted much of my effort to modeling the impact of non-pharmaceutical interventions for mitigating disease transmission. Infectious diseases spread within a social contact network, and many strategies for mitigating spread can be thought of in terms of this network (e.g., social distancing and contact tracing). A framework that incorporates realistic contact networks is essential to explore the efficacy of such interventions. I developed a flexible stochastic network model that incorporates the structure of data-driven contact networks and other extensions that are critical for studying COVID-19 (e.g., pre- and asymptomatic transmission, hospitalization, testing, tracing, isolation). This work led to collaborations with researchers and stakeholders across disciplines. In Part 2, I describe how we used the model framework I developed to inform model-driven mitigations for schools and workplaces.