COMPUTATIONAL MODEL OF SYNAPTIC DEVELOPMENT

点击翻译按钮获取中文摘要

Because developing cortical response properties are highly modifiable as a function of experience and because synaptogenesis is a fundamental process of development, we are studying the possible roles synaptogenesis might play in such modifications of response properties. Because ocular dominance is the response property that has received the most experimental attention, our developmental models simulate the various experimental manipulations that do, or do not, produce changes of this response for layer IV cells of area 17 of the cerebral Cortex. (ocular dominance is the technical term describing the relative cell firing a single cell produces in response to a stimulus given to just the left eye versus the same stimulus given to just the right eye. If a cell fires equally fast to either presentation, it is described as binocular; if the cell fires more for one eye than the other, the cell is described as monocular and preferring the eye associated with the larger response.) Our goal is to find the form of the adaptive processes, i.e. synaptogenesis, synaptic shedding, and associative synaptic potentiation and depression, that are compatible with each other and that together imply a model of the experimental observations. Such models are tested with computer simulations. Such simulations are useful, not only for testing theories of visual development, but also for bringing into focus issues that are not intuitively available to experimentalists. Such models increase both the efficiency of experimental research and our understanding of brain development. So far we have simulated developing ocular dominance under four conditions: a normal environment; monocular deprivation; and the effect of two drug treatments on monocular deprivation. The simulations of these conditions show how the response properties of individual cells change as a function of time. The results of this work encourage us because we find that the desired trends in response properties were achievable without making the model overly complex and because steady-state response properties, without undue oscillations, were achievable even after we altered input statistics and/or simulated drug manipulations. We now propose to simulate many more cells so we can create ocular dominance histograms. Histograms are needed because the dichotomy of monocular versus binocular is always quantitiveIy refined into degrees of monocular or binocular responsiveness, i.e. the published data are in the form of such histograms. In trying to understand the simplest model that fits the published histograms, we will investigate a variety of variables that might destroy the performance of the model or, more likely, increase the robustness of the model. Such variables include the formulation of the synaptic modification equations, the role of threshold functions, the role of inhibition, etc. Also part of the investigation is the interaction of the input statistics; i.e. the characteristics of lateral geniculate activity will be varied, and as they vary, they will interact with the variables just mentioned and interact with the size of the network.