Real-time mapping and adaptive testing for neural population hypotheses

神经群体假设的实时映射和自适应测试

• 批准号: 10838394
• 负责人: John Pearson
• 金额: $ 20.11万
• 依托单位: DUKE UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-09-15 至 2025-09-14
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10838394
• 关键词: Algorithms; Animals; Behavior; Behavioral; Benchmarking; Biological Models; Brain; Calcium; Case Study; Cells; Complex; Computer software; Data; Data Analyses; Data Set; Dimensions; Documentation; Educational workshop; Electrophysiology (science); Ensure; Experimental Designs; Explosion; Frequencies; Goals; ImProv; Image; Individual; Intervention; Link; Maps; Methods; Modeling; Neurons; Neurosciences; Nonlinear Dynamics; Population; Population Dynamics; Process; Property; Recurrence; Research Personnel; Running; Sampling; Series; Signal Transduction; Software Tools; Sorting; Specific qualifier value; Stimulus; System; Techniques; Technology; Testing; Time; Visualization; Work; Zebrafish; autoencoder; complex data; conditioning; data pipeline; design; dynamic system; experimental study; graphical user interface; improved; insight; interest; light weight; model organism; neural; neural model; open source; optogenetics; parent grant; predictive modeling; simulation; tool; transfer learning; two-photon; usability; virtual

ABSTRACT Recent advances in neural recording technologies have made it possible to study increasingly large and di- verse subsets of neurons, producing a growing interest in the collective computational properties of neural pop- ulations. Ideally, causally testing these population hypotheses requires timing and selecting experimental ma- nipulations based on the current state of neural dynamics, but technical limitations have rendered this difficult in practice. However, recent work on real-time preprocessing and modeling of neural data has demonstrated that up-to-the minute estimates of neural population dynamics are indeed possible, opening the door to adap-tive experiments in which the design of the task changes based on incoming data. The goal of this proposal is to disseminate these advances to the widest possible audience of systems neuroscientists by: 1) Designing and validating new methods for mapping neural states and behavior online. 2) Developing algorithms for optimally timing and selecting experimental manipulations based on these instantaneous neural and behavioral states. 3) Making improv, our platform for adaptive experiments, easier to install, use, and configure for diverse model organisms and hardware setups. By allowing researchers to test ideas online, such tools will facilitate rapid “drill-down” from the whole brain to the local circuit levels, maximizing statistical efficiency in limited experi- mental time and providing stronger causal inferences for neural population hypotheses, with broad implications for systems neuroscience. 1

抽象的 神经记录技术的最新进展使得研究越来越大和越来越分散成为可能。 神经元的子集，使人们对神经元的集体计算特性越来越感兴趣 的规定。理想情况下，因果检验这些群体假设需要时机和选择实验材料 基于神经动力学当前状态的操纵，但技术限制使得这变得困难 实践。然而，最近关于神经数据实时预处理和建模的工作表明： 对神经群体动态的最新估计确实是可能的，这为自适应性打开了大门 任务设计根据传入数据而变化的实验。该提案的目标是 通过以下方式向最广泛的系统神经科学家传播这些进展：1）设计和 验证在线映射神经状态和行为的新方法。 2）开发优化算法 根据这些瞬时神经和行为状态来计时和选择实验操作。 3） 让我们的自适应实验平台“improv”更易于安装、使用和配置以适应不同的模型 有机体和硬件设置。通过允许研究人员在线测试想法，此类工具将促进快速 从整个大脑“深入”到局部回路水平，在有限的经验中最大限度地提高统计效率 心理时间并为神经群体假设提供更强的因果推论，具有广泛的影响 用于系统神经科学。 1