Understanding the constraints on sex ratio adaptation using artificial neural networks

How perfect should behaviour be? There is little doubt that, given sufficient time, and genetic variation, natural selection can produce organisms displaying startlingly precise adaptations to their surroundings. At the same time, a variety of processes can constrain the ability of populations to reach adaptive peaks. Determining the relative importance of these processes is a major challenge still facing evolutionary biology. We will work on a model system that will allow us to say how obtaining and processing information can limit adaptation. We will work on parasitic wasps, which can choose the sex of their offspring by deciding whether or not to fertilize their eggs. The choice of offspring sex in this group has been the subject of much previous work, and the wasps generally show a good but imperfect fit to predicted behaviour. We will ask if we can better understand the behaviour shown in this group of organisms by incorporating the limitations of sensory and nervous systems into our predictions using a modeling approach known as artificial neural networks. The new predictions will then be tested against existing datasets and against new data, obtained by manipulating the information available to a parasitic wasp in the laboratory.

How perfect should behaviour be? There is little doubt that, given sufficient time, and genetic variation, natural selection can produce organisms displaying startlingly precise adaptations to their surroundings. At the same time, a variety of processes can constrain the ability of populations to reach adaptive peaks. Determining the relative importance of these processes is a major challenge still facing evolutionary biology. We will work on a model system that will allow us to determine how obtaining and processing information can limit adaptation. Our system is sex ratio evolution in parasitic wasps. The study of sex ratios in parasitic wasps is one of the few areas of evolutionary biology where we can expect, and have observed, a quantitative fit between theory and data. This means that studies of sex allocation in parasitic wasps can be used as a model trait (tool) for studying the relative power of adaptation and constraint. Our modelling approach will be to incorporate into sex ratio theory explicit assumptions about information acquisition and processing, using artificial neural networks. Artificial neural networks are simple analogues of real neural processing systems that allow us to embody well defined mathematical relationships but do so in a way that retains the special biases and properties of real neural processing. We will use such models to explain some outstanding problems in sex ratio evolution: a) why, in solitary wasps, the switch from male to female offspring with increasing host size is gradual rather than sudden as predicted by simple models b) why female fig wasps show a better match to simple model predictions about their sex ratio for foundress numbers that they more commonly encounter in nature c) whether deviations from simple sex ratio predictions in the wasp Nasonia vitripennis are the result of its use of an indirect cue to assess foundress number.