NONLINEAR MODEL OF HIPPOCAMPUS

在过去的一年里，我们在发展 并将几种基本的建模方法应用于 点过程输入神经元的输入/输出或非参数模型 海马体等系统。一是坚持发展。 以及分解的自适应估计程序的应用 海马体的研究。使用这种方法，我们已经能够 成功地估计了负数的线性和非线性性质 给出反馈元素(类似于抑制性中间神经元) 仅针对存在的馈通元件的输入/输出数据 没有抑制，即没有来自 反馈途径。第二，我们开发了一种新的方法来 用人工神经网络估计神经元的输入/输出特性 (前馈)神经网络，我们已经演示了 特别适用于估计高阶非线性。 电生理数据。我们特别兴奋的是 扩展这种估计高阶的方法的潜力 多输入条件的交叉核。第三，第一次 我们已经将我们的非线性系统方法应用于 细胞内记录突触电位。我们最近做了 对以下组件的开环和开环AMPA和NMDA组件建模 谷氨酸能突触传入海马神经元，衍生出一种 NMDA正反馈的动力学表示 受体通道，并能准确预测控制EPSPS的范围 输入模式的范围。此外，该模型还可以解释 N-甲基-D-天冬氨酸受体介导的闭环和开环的差异 突触反应。这是研究最广泛的两种受体 子类型，因为它们相互作用的动态决定了方向 和活性依赖的突触可塑性的大小 海马体、新皮质和其他大脑区域。最后，我们有 在开发高效电路方面取得了长足的进步 海马区核心模型的模拟VLSI实现设计 神经元，一个应该为发展提供手段的方向 用于模拟全局海马区的大规模并行体系结构 系统属性。

During this past year we have made major progress on developing and applying several of the essential modeling methodologies for input/output, or nonparametric, models of point-process input neural systems such as the hippocampus. First, we have pursued development and application of adaptive estimation procedures for decomposition studies of the hippocampus. Using this approach, we have been able to successfully estimate linear and nonlinear properties of negative feedback elements (analogous to inhibitory interneurons) given input/output data for only the feedthrough element in the presence and absence of inhibition, i.e., without direct observations from the feedback pathway. Second, we have developed a new means for estimating input/output properties of neural elements using artificial (feedforward) neural networks, which we have demonstrated is particularly powerful for estimating higher order nonlinearities from electrophysiological data. We are particularly excited about the potential for expanding this approach for estimating higher order cross-kernels for multiple-input conditions. Third, for the first time we have applied our nonlinear systems approach at the level of intracellularly recorded synaptic potentials. We have recently modeled the closed- and open-loop AMPA and NMDA components of glutamatergic synaptic input to hippocampal neurons, have derived a representation of the dynamics of the positive feedback to the NMDA receptor-channel, and can accurately predict control EPSPs to a wide range of input patterns. In addition, the model can account for differences between closed- and open-loop NMDA receptor-mediated synaptic responses. These are two of the most widely studied receptor subtypes, as the dynamics of their interaction determine the direction and magnitude of activity-dependent synaptic plasticity in hippocampus, neocortex, and other brain regions. Finally, we have made considerable progress in developing highly efficient circuit designs for analog VLSI implementation of kernel models of hippocampal neurons, a direction that should provide the means for developing massively parallel architectures for simulating global hippocampal system properties.