Collaborative Research: Spatially Distributed Computation in a Small Neural Network

9604505 Hartline 长期以来，神经生物学家一直将神经细胞视为神经系统功能的基本单位。 在这种观点中，细胞（或“神经元”）接收来自选定来源（感觉细胞、其他神经细胞、血液中的激素）的输入，处理这些输入整体中包含的信息，并向其他神经元或肌肉发送代表处理的统一结果的信号。 然而，越来越明显的是，单个神经元的这种功能统一在某些情况下（也许在许多情况下）可能是不正确的。 相反，一个单元的不同部分可以分别基于对整个单元接收的输入的选定“本地”采样的处理来发出自己的信号。 这项研究将评估通过小型神经网络在细胞中进行这种“本地”计算的可能性。 该网络特别适合这项研究，因为它具有足够少的神经元，可以对输入到特定目标神经元的所有细胞的活动进行生理监测。 经验一再表明，在这个简单的模型系统（来自螃蟹的神经节，称为“口胃神经节”）中更容易学习的原理可以应用于更复杂的系统，包括哺乳动物和人类的神经系统。 单个神经元将被注射荧光染料，使它们在“共焦”显微镜下可见。 共焦显微镜将记录来自每个细胞拥有的大量分支延伸的荧光，从而允许准确重建细胞的复杂形状。 由此，以及在注射荧光染料的同时进行一些简单的电测量，将构建一个计算机模型，预测电压信号通过细胞不同部分的传播。 如果在单元中的任何一点生成的信号在传播到单元的其他部分时变化相对较小，则这将证明单元中几乎不存在“本地”计算的可能性，因为所有部分都将接收相同的处理后的信号。 正如初步数据表明的那样，如果信号在从单元中的一个点传递到另一个点的过程中可以发生显着改变，那么本地计算的可能性就会大得多。 许多研究都集中在本地计算能力对各种生理因素的敏感程度。 例如，随着其他神经细胞的输入变得更强，目标细胞的电特性会发生变化，因此它更有可能分解成小的局部计算区域。 另一方面，如果来自其他细胞的输入广泛分布在目标神经元的表面上，则目标细胞可以以更统一的方式起作用。 该项目正在研究这些不同的参数如何影响真实神经元的计算风格，以及单个神经元的计算风格如何随着神经系统活动状态的变化而变化。

9604505 Hartline For a long time, neurobiologists have viewed the nerve cell as being the fundamental unit of function in the nervous system. In this view, a cell (or "neuron") receives inputs from selected sources (sensory cells, other nerve cells, hormones in the blood), processes the information contained in the totality of these inputs and sends out a signal to other neurons or to muscles that represent a unified result of the processing. However, it is becoming increasingly apparent that this unity of function of the single neuron may in some cases, and perhaps in many, be incorrect. Instead, different parts of one cell may each be sending out its own signal based on processing of a selected "local" sampling of inputs received by the whole cell. This research will assess the possibilities for such "local" computation in cells from a small neural network. This network is particularly suited to the research because it has few enough neurons that activity from all cells having input onto a particular target neuron can be monitored physiologically. Experience has shown repeatedly that principles learned more easily in this simple model system (a ganglion from a crab termed the "stomatogastric ganglion"), can be applied to more complex systems, including the nervous system of mammals and man. Single neurons will be injected with a fluorescent dye that makes them visible in a "confocal" microscope. The confocal microscope will record the fluorescence coming from the multitude of branching extensions possessed by each cell and thus will allow accurate reconstruction of the complex shape of the cell. From this, and from some simple electrical measurements that can be made at the same time the fluorescent dye is injected, a computer model will be constructed of the predicted spread of voltage signals through the different parts of the cell. If the signals generated at any one point in the cell are changed relatively little in traveling to other parts of the cell, this will be evidence that there is little likelihood of "local" computation in the cell, since all parts will receive the same processed signal. If, as preliminary data suggest, signals can be significantly altered in passing from one point in a cell to another, the likelihood of local computation is much greater. Much of the research focuses on how sensitive the local computation capabilities are to various physiological factors. For example, as input from other nerve cells becomes stronger, the electrical properties of the target cell change so that it is more likely to break up into small local computational regions. On the other hand, if the input from other cells is broadly distributed over the surface of a target neuron, the target cell may act in a more unified fashion. The project is investigating how such different parameters influence the computational style of real neurons and how computational styles of a single neuron may change as the state of activity of the nervous system changes.