Network and dendritic mechanisms of context-dependent cortical computation

包括精神分裂症、帕金森病和阿尔茨海默病在内的许多疾病正在细胞和分子水平上得到深入研究，在它们如何改变突触传递和单个神经元功能方面取得了相当大的进展。然而，细胞水平的神经元功能如何转化为大规模网络功能和认知的破坏尚不清楚。细胞功能和行为之间的关键联系是神经元树突中突触输入的整合，这可以为每个神经元提供自己的计算机制，从而形成神经元网络执行的计算。由于两个主要障碍，研究亚细胞树突功能如何影响大脑功能和行为，特别是通过参与高级认知的联想皮层，一直具有挑战性。首先，在啮齿类动物中，参与联想皮层的行为与头部固定不相容，限制了我们在复杂行为中测量树突信号的能力。其次，解释来自自然行为的数据在计算上和概念上都很困难：感觉编码可以通过平均数百次重复的试验来理解，但以这种方式研究决策使动物记住简单而稳定的刺激-反应关联，这不需要以同样的方式联想皮层。本提案将使用克服这些问题的新方法来检查树突和网络计算在复杂联想行为中的贡献。这个项目利用了一种新的方法来对自由行为进行细胞水平的成像，这项任务需要动物进行联想计算，但很容易训练。该候选人具有系统神经科学、电生理学、钙成像和工程学方面的专业知识，为本项目提供了坚实的基础。K99将提供树突生理学方面的训练，以及将复杂任务与个体神经元和神经元网络的活动联系起来所需的计算方法，尽管行为存在可变性。主要导师Mark Harnett博士将提供树突功能和亚细胞计算研究方法方面的指导。共同导师Ila Fiete博士将提供与该任务相关的计算方法培训。Mehrdad Jazayeri博士对人口动态建模和Cengiz Pehlevan博士对基础计算的数学特性的额外建议将确保所有相关科学方法的高水平培训。该培训将使候选人能够完成这些目标，并开发一个独立的研究计划，专注于复杂自然行为背后的神经计算。这项研究的结果将为细胞计算如何促进复杂行为提供新的见解，这对于理解在经典啮齿动物行为中不易研究的具有重要行为表型的疾病，如痴呆和精神分裂症，将是重要的。

Many disorders including schizophrenia, Parkinson's disease, and Alzheimer's disease are being studied in-depth on the cellular and molecular level, and considerable progress has been made on how they change synaptic transmission and single neuron function. However, how cellular-level neuronal function translates to disruptions in larger-scale network function and cognition is not well understood. A key link between cellular function and behavior is the integration of synaptic inputs in the dendrites of neurons, which can give each individual neuron its own computational mechanisms that in turn shape the computations performed by networks of neurons. Efforts to study how sub-cellular dendritic function contributes to brain function and behavior, especially via the associative cortices involved in higher-level cognition, have been challenging, due to two major barriers. First, in rodents, the behaviors that engage associative cortices are incompatible with head fixation, limiting our ability to measure dendritic signals during complex behavior. Second, interpreting data from natural behaviors is computationally and conceptually hard: sensory encoding can be understood by averaging hundreds of trial repetitions, but studying decision-making in this manner makes animals memorize simple and stable stimulus-response associations, which do not require associative cortices in the same manner. This proposal will use new approaches that overcome these issues to examine the contributions of dendritic and network computation in a complex associative behavior. This project exploits a novel approach for cellular-level imaging in free behavior, and a task that requires animals to perform associative computations but is easy to train. The candidate has expertise in systems neuroscience, electrophysiology, calcium imaging, and engineering, providing a strong foundation for this program. The K99 will provide training in dendritic physiology, and the computational methods required for linking complex tasks to the activity of individual neurons and networks of neurons despite variability in behavior. The primary mentor Dr. Mark Harnett will provide mentorship in dendritic function and methods for studying subcellular computation. The co- mentor Dr. Ila Fiete will provide training in computational methods relevant to the task. Additional advice from Dr. Mehrdad Jazayeri on modeling population dynamics and Dr. Cengiz Pehlevan on the mathematical properties of the underlying computations will ensure high-level training in all relevant scientific approaches. This training will enable the candidate to complete these aims and to develop an independent research program focused on neural computations underlying complex natural behaviors. The results of this research will provide new insight into how cellular computations contribute to complex behaviors, which will be important for understanding disorders with non-trivial behavioral phenotypes that are not easily studied in classical rodent behavior, such as dementia and schizophrenia.