The cellular and molecular basis of motion computations

神经网络如何执行许多关键计算尚不清楚。在感觉系统中，各种计算从环境中提取显著信息并指导适当的行为。尽管经过数十年的研究，我们对这些过程的理解仍然是支离破碎的：在某些系统中，已经确定了对不同感官线索做出反应的特定神经元;在其他系统中，特定的行为输出或预测生理或行为的计算模型是已知的。然而，对神经元如何获得特定的生理特性，它们如何在电路中组织以及这些电路如何引导不同的行为的完整理解尚未在任何系统中建立。全面了解大脑各个层面的功能将为治疗精神或神经疾病开辟新的途径。从昆虫到人类的动物都使用视觉信息，特别是运动线索，在环境中导航，捕捉猎物或逃避捕食者。由于运动视觉需要电路来整合空间和时间上的视觉信息，因此长期以来一直被认为是理解大脑功能的范例计算。描述如何提取运动信息的计算模型已经存在了50多年，并解释了从苍蝇到人类的运动感知。然而，实现这些模型的神经回路在很大程度上仍然未知。此外，许多分子和细胞机制调节突触活动或调节已识别神经元的细胞特性，但它们很少与特定的行为相关计算相关。虽然目标应该是将这些机制与最终的读出，动物行为联系起来，但这在许多系统中是不可能的。我打算通过研究遗传模式生物果蝇的运动检测来实现这一目标。在苍蝇中，运动引导的行为已经被详细研究和计算描述。我提出了细胞生物学和遗传学的方法来操纵运动检测电路中的关键神经元。结合生理学和定量行为分析，这些研究将确定指导运动行为反应的细胞和分子机制。然后，我将应用我构建的强大的遗传工具包来操纵特定神经元中的神经元功能，从而识别特定运动计算的基础电路。最后，我将定义一个大类的视觉系统神经元的响应空间，并确定其在处理流的功能。总之，这个项目将提供一个集成的理解，一个范例的神经计算，链接到一个定义的电路和电机输出的分子机制。这项工作将确定神经系统可以垂直整合分子，细胞和电路机制来计算特定输入的行为关键输出的一般机制。

How neural networks perform many critical computations is not understood. In sensory systems, a variety of computations extract salient information from the environment and guide appropriate behavior. Despite decades of work, our understanding of these processes remains fragmentary: in some systems, specific neurons have been identified that respond to distinct sensory cues; in others, specific behavioral outputs or computational models that predict physiology or behavior are known. However, a complete understanding of how neurons gain specific physiological properties, how they are organized in circuits and how these circuits guide distinct behaviors, has not been established in any system. A comprehensive understanding of brain function at all levels will open up new avenues for treating psychiatric or neurological diseases.Animals ranging from insects to humans use visual information, especially motion cues, to navigate through the environment, capture prey, or escape predators. Because motion vision requires circuits to integrate visual information over both space and time it has long been considered a paradigmatic computation for understanding brain function. Computational models that describe how motion information can be extracted have existed for more than 50 years, and explain motion perception from flies to humans. However, the neural circuits that implement these models remain largely unknown. Moreover, many molecular and cellular mechanisms regulate synaptic activity or modulate cellular properties in identified neurons, but they have only rarely been associated with specific, behaviorally relevant computations. While the goal should be to link such mechanisms to the ultimate read-out, animal behavior, this is impossible in many systems. I intend to achieve this by studying motion detection in a genetic model organism, the fruit fly Drosophila. In flies, motion-guided behaviors have been studied in detail and described computationally. I propose cell biological and genetic approaches to manipulate critical neurons in motion detecting circuits. In combination with physiology and quantitative behavioral analysis, these studies will identify the cellular and molecular mechanisms that guide behavioral responses to motion. I will then apply a powerful genetic toolkit I have built to manipulate neuronal function in specific neurons, thereby identifying the circuits that underlie specific motion computations. Finally, I will define the response space of a large class of visual system neurons and identify their function in processing streams.In summary, this project will provide an integrated understanding of a paradigmatic neural computation that links molecular mechanisms to a defined circuit and to motor output. This work will identify general mechanisms by which a nervous system can vertically integrate molecular, cellular and circuit mechanisms to compute behaviorally critical outputs from specific inputs.