Compartmental models and genetic algorithms provide insight into structure-function relationships in the lateral dendrite of the Mauthner neuron.

摘要

Nerve cells use specialized geometries and electrical properties to effect computational transformation of their inputs. In real systems, it is possible to observe the behavioral, and often the electrical, result of these computations, but often impossible to observe their mechanism. Detailed mathematical models provide a system in which many of the properties of the real system are maintained, but the internal state and mechanism of the system is rendered easily accessible. I have used such a model to investigate the potential for a single cell computation of sound source location by the Mauthner cell of goldfish, and have used this model to predict the distribution of input synapses on the cell. In addition, I have developed several techniques for investigating the relationship between structure and function in computing interneurons. The first of these is to make large numbers of small variations to the structure of a realistic model, and record the effects of these changes on computational function. The second is to use genetic optimization algorithms to optimize a population of randomly structured neurological models for a particular function, and subsequently determine by variance analysis what structural features were constrained by this functional requirement.; My Mauthner cell model indicates that the Mauthner cell contains all the structural properties needed for single cell computation of sound source location. It predicts that, if this computation is, in fact, occurring, input synapses representing different modalities of a sound stimulus will be grouped into clusters on the dendrite, with greater spacing between clusters representing the same phase of pressure and acceleration modalities than between clusters representing opposite phases.; The genetic optimization also indicates that spacing of inputs is essential to the sound localization computation. In addition, it provides the result that several structural traits present in the real Mauthner cell occur in optimized members of the population. Since the initial population is structurally random, these structures are presumably a result of constraining for cells that succeed at a computational task similar to that of the Mauthner system.