Computer Evolution of Chemotaxis in Model Nematodes

tration. Since each segment follows the one in front of it during sinusoidal locomotion, the nematode body is represented by head location and velocity, with head direction controlled by the action of 2 opposing neck muscles. Finally, the model neural circuit consists of 2 chemosensory neurons and 2 neck muscle neurons, each representing identified classes of neurons in C. elegans , whose synaptic connections are consistent with the known anatomical connectivity. This circuit is driven by a sinusoidal oscillation representing the nematode locomotion system. In order to set the electrophysiological parameters of the neural circuit model, an evolutionary algorithm was used. An evolutionary algorithm is a search technique loosely based on biological evolution [Mitchell, 1998]. A search begins with a random population of genetic strings that encode the unknown phenotypic characteristics of interest (here, the parameters of the neural circuit model). Each individual is then assigned a fitness based on the quality of its performance on the given task (here, the chemotactic success of the entire brain-body-environment model averaged over multiple trials). Individuals are then selected to serve as parents for the next generation with a probability related to their fitness. From the selected parents, a new generation of children are then produced by randomly swapping portions of 2 parents (crossover) and making small modifications (mutation). Once a new The nematode worm Caenorhabditis elegans is a remarkable organism for behavioral neuroscience. Not only does this millimeter-long soil-living animal exhibit a rich behavioral repertoire (including withdrawal reflexes, crawling, swimming, a variety of taxes, social feeding, egg-laying, habituation and associative conditioning), but it does so with a nervous system containing only 302 neurons, for which the anatomical connectivity is completely known from serial section electron microscopy [White et al., 1986]. Despite the small size and pressurized body of C. elegans , progress is also being made on electrophysiological analysis of its nervous system using genetic and developmental manipulations, optical imaging, and whole-cell patch-clamp techniques [Schafer, 2006]. However, we are still a very long way from knowing even which of the over 6,000 synaptic connections that make up this nervous system are excitatory or inhibitory. A recent paper by Izquierdo and Lockery [2010] in the Journal of Neuroscience demonstrates a new way to use computational techniques to compensate for missing electrophysiological data. The paper focuses on a model of salt klinotaxis, a recently-discovered form of chemotaxis in C. elegans in which orientation is continuously adjusted along a line of steepest ascent in a salt gradient [Iino and Yoshida, 2009]. The environment is modeled as a static cone-shaped gradient of salt concenPublished online: December 24, 2010