Testing a computational model of neural responses in autism

点击翻译按钮获取中文摘要

PROJECT SUMMARY/ABSTRACT This proposal will test a novel, computationally-motivated hypothesis about neural dysfunction in autism spectrum disorder (ASD). ASD is a heterogeneous neurodevelopmental disorder of unknown etiology. However, a unifying theme of numerous proposals is that there is a pervasive disruption of neural excitatory/inhibitory (E/I) balance. A major limitation of the E/I hypothesis is that it describes a property of individual neurons; how that property scales up to neural circuits and how it relates to behavior – the level at which ASD is described – is not well specified. Neural computational models offer a way to bridge the divide between single-unit properties and behavior, and bring the necessary specificity to test possible changes in E/I in ASD. One well-established neural computation that directly relates to E/I is “divisive normalization”, a computational framework that characterizes neural responses as the ratio of net excitatory relative to net suppressive input. Here we aim to test the hypothesis that ASD involves disrupted divisive normalization using vision as a model system. We will test two possible mechanisms of weakened divisive normalization. The first is the traditionally posited disruption of local, within-area circuits that mediate suppressive drive. The second is a novel hypothesis based on recent empirical findings in our lab. We have shown enhanced suppressive feedback of responses from higher stages to lower stages of visual processing in individuals with ASD. We suggest this enhanced suppressive feedback reduces responses of neurons that would otherwise participate in divisive normalization. This hypothesis makes specific predictions about the conditions under which disrupted divisive normalization will be observed in ASD. We will test these predictions using a combination of functional MRI, ERP, and diffusion MRI.