Files

Abstract

Adapting populations have often been characterized in the context of static environments. However, many natural environments vary on timescales similar to or faster than the rate of evolutionary adaptation. Examples of time-varying environments include the environment faced by the adaptive immune system when generating antibodies against highly mutagenic viruses or the environment faced by an organism attempting to escape from an incoming predator. We begin by exploring adaptation against HIV, a virus that mutates on the same timescale as it takes for B-cells to evolve antibodies. In our study, we propose a conceptual framework in which there exists generalist and specialist phenotypes. Specialist phenotypes are strategies that confer high fitness in a given environment, and may not confer high fitness in any other environment. Generalist phenotypes, on the other hand, are 'jack-of-all-trades' strategies and work well across a family of environments, though they may not work as well as a specialist phenotype for a given environment. We are able to demonstrate that the preferred phenotype depends on the variation of the environmental landscape. Further, we identify that generalist phenotypes confer fitness by exploiting the correlation structure of the time-varying environment. We then consider the sensory encoding scheme used by the retina against changing visual scenes. Unlike in the previous study, the natural environment changes much more rapidly than the evolutionary adaptation time of the sensory encoding scheme. Consequently, we explore how a sensory encoding scheme can be predictive for a fixed level of compression. Using the information bottleneck method, we explore the optimal sensory encoding schemes for a range of time-varying environments relevant to the visual system. In addition, we also explore the transferability of a sensory encoding scheme, identifying the best schemes when the autocorrelation structure of the environment itself varies.

Details

Actions

PDF

from
to
Export
Download Full History