Neural Representations of Abstract Sequences

项目摘要 研究：拟议的研究将调查抽象序列的神经表征。摘要 序列由在感觉刺激上概括的高阶模式定义（例如，AAAB，&*）。 序列处理缺陷出现在一系列神经精神疾病中，包括强迫症 强迫症（OCD）、帕金森病和额叶功能障碍。这些赤字的来源很难确定。 确定，因为它们背后的神经机制仍然知之甚少。以前的工作表明， 外侧前额叶皮层（LPFC）参与序列加工，但抽象的神经表征 序列尚未被研究。拟议的研究解决了这一知识差距。我专注于两个 问题：序列表示是否抽象，顺序行为是否需要多个 神经亚群，以促进灵活的编码。我用两个互补的方法来处理这些问题。 方法：非人灵长类动物的神经记录和递归神经网络（RNNs）建模。在Aim中 1，我将使用fMRI引导的神经记录来检验猕猴LPFC中序列表征的假设， 是独立的刺激身份，导致跨刺激和任务上下文的概括性。在目标2中，我将 使用低秩RNN来测试序列监测是否需要比非序列监测更多的神经元亚群。 顺序延迟匹配到样本任务，灵活的刺激-反应映射的签名。 这些研究将扩展我们对抽象序列表示的认识。此外，这些结果还可作为 案例研究，以了解泛化的两个关键特征：表征稳定性和实现性 灵活性.结合我们小组的其他工作，这项研究的数据将在以下方面建立一座桥梁： 灵长类动物电生理学，灵长类动物fMRI和人类fMRI，为跨物种模型的发展提供信息 人类疾病。 环境与培训：我的环境非常适合拟议的培训。利用他们的专业知识， 我的赞助商和共同赞助商将在多电极记录，大型神经数据集的分析， 使用RNN进行计算建模。此外，我的赞助商领导着为数不多的进行功能磁共振成像的实验室之一， 在清醒的猴子中进行的实验，使我们能够在功能上靶向感兴趣的区域， 录制.布朗大学的合作研究社区、研究设施和计算资源 进一步支持我的培训计划。除了我的研究，我的培训将包括神经系统的研讨会， 分析和建模，参加科学会议，专业发展和科学培训 沟通，指导，并在研究负责任的行为。本次培训将提供理想的 我的职业目标是结合计算和系统神经科学来研究泛化， 我自己的实验室

PROJECT SUMMARY Research: The proposed research will investigate the neural representation of abstract sequences. Abstract sequences are defined by higher-order patterns that generalize across sensory stimuli (e.g., AAAB, &&&*). Deficits in sequence processing arise in a range of neuropsychiatric disorders, including Obsessive-Compulsive Disorder (OCD), Parkinson’s disease, and frontal lobe dysfunction. The source of these deficits is difficult to determine, as the neural mechanisms behind them remain poorly understood. Previous work suggests that lateral prefrontal cortex (LPFC) contributes to sequence processing, but the neural representation of abstract sequences has not been investigated. The proposed studies address this knowledge gap. I focus on two questions: whether sequence representations are abstract, and whether sequential behavior requires multiple neural subpopulations to facilitate flexible coding. I approach these questions using two complementary methods: neural recording in nonhuman primates and modeling with recurrent neural networks (RNNs). In Aim 1, I will use fMRI-guided neural recording to test the hypothesis that sequence representation in macaque LPFC is independent of stimulus identity, leading to generalizability across stimuli and task contexts. In Aim 2, I will use low-rank RNNs to test whether sequence monitoring requires more neuronal subpopulations than a non- sequential delayed-match to sample task, a signature of flexible stimulus-response mapping. These studies will expand our knowledge of abstract sequence representation. Moreover, the results serve as a case study to understand two key features of generalization: representational stability and implementational flexibility. In combination with other work from our group, the data from this study will create a bridge between primate electrophysiology, primate fMRI, and human fMRI, informing the development of cross-species models of human disease. Environment & Training: My environment is ideally suited for the proposed training. Drawing on their expertise, my Sponsor and Co-Sponsor will train me in multi-electrode recording, analysis of large neural datasets, and computational modeling using RNNs. Furthermore, my Sponsor leads one of the few labs conducting fMRI experiments in awake monkeys, giving us the unique ability to functionally target regions of interest for neural recording. The collaborative research community, research facilities, and computational resources at Brown further support my proposed training. In addition to my research, my training will include workshops in neural analysis and modeling, participation in scientific meetings, professional development, and training in scientific communication, mentorship, and responsible conduct in research. The proposed training will provide ideal preparation for my career goal of combining computational and systems neuroscience to study generalization in my own lab.