CRCNS: Unsupervised Learning of Hippocampal Sequence Dynamic in Sleep

CRCNS：睡眠中海马序列动态的无监督学习

基本信息

批准号：
10614754
负责人：
KAMRAN DIBA
金额：
$ 7.06万
依托单位：
RICE UNIVERSITY
依托单位国家：
美国
项目类别：
财政年份：
2019
资助国家：
美国
起止时间：
2019-08-15 至 2024-05-31
项目状态：
已结题

项目摘要

In unit recordings from large populations of neurons, fast compressed sequential firing of neurons during rest and early sleep have been found to replay patterns first observed in active awake experience. These remarkable patterns have sparked widespread interest in the scientific community and beyond. Sequence replay is now considered to play a critical role in the long-term stabilization and storage of mnemonically important information. However, despite the general acknowledgement of the importance of the sequential structure, very little is known about the null background against which replay is compared. Specifically, are apparently 'non-replaying' spike patterns, as seen in late sleep, just simply noise? Because replay is typically assessed by comparison against a fixed known template, most methods can only determine whether the resemblance to the template is more than what might be expected from random spike trains. But these methods cannot appraise whether other patterns remain in the nonsignificant events. Recently, the Diba and Kemere labs successfully collaborated to address precisely this issue. We developed methods based on hidden Markov models (HMMs) to uncover temporal structure in spike trains of neurons in an unsupervised template-free manner. In this proposal, we aim to further improve these methods and to evaluate the hidden structure of spike trains in hippocampal neuronal populations during sleep. In our second specific aim, we will use HMMs to determine both co-active ensemble ("contextual") and temporal patterns ("sequential") structure in hippocampal spike trains in both pre- and post-task sleep. In the third specific aim, we will probe the essence of sleep replay further, by exposing animals to multiple novel and familiar maze environments prior to long durations of sleep. In the fourth specific aim, we will perform closed-loop disruption of neuronal population patterns to examine the causal interplay and reverberation of these patterns from early to late sleep. In summary, our proposal is designed to provide strongest characterization to date of the structure of "noise" in replay events. RELEVANCE (See instructions): This study will provide an opening to evaluate the role of sleep in reorganizing information in the brain and help to identify critical time windows and neuronal activities during sleep which are particularly important for information storage and stabilization. Our assumptions and deductions about the nature and purpose of sleep implicitly inform all manner of public policy, from the durations of shifts for hospital and relief workers, to morning start times of public schools. Understanding the function and mechanisms of sleep H

在大量神经元种群的单位记录中，在已经发现休息和早期睡眠是在主动清醒体验中首先观察到的模式。这些出色的模式引发了对科学界及其他地区的普遍兴趣。顺序现在，重播被认为在长期稳定和储存中起着至关重要的作用重要信息。然而，尽管一般承认顺序的重要性关于比较重播的无效背景的结构，几乎没有什么了解。具体而言，显然是“非重新播放”的尖峰图案，如晚睡觉时所见，只是噪音吗？由于通常通过与固定已知模板进行比较来评估重播，因此大多数方法可以仅确定与模板的相似性是否超出随机尖峰火车。但是这些方法无法评估其他模式是否保留在不重要的事件。最近，Diba和Kemere Labs成功合作，准确地解决这个问题。我们开发了基于隐藏的马尔可夫模型（HMM）的方法以发现时间无监督模板的方式的神经元尖峰列车中的结构。在此提案中，我们的目标是进一步改进这些方法并评估海马中尖峰列车的隐藏结构睡眠期间的神经元种群。在我们的第二个特定目标中，我们将使用HMM来确定海马尖峰中的共同活性集合（“上下文”）和时间模式（“顺序”）结构在任务前和任务后睡觉。在第三个特定目标中，我们将探究睡眠重播的本质此外，通过在长时间之前将动物暴露于多个新颖和熟悉的迷宫环境睡觉。在第四个特定目标中，我们将对神经元种群模式进行闭环破坏至检查这些模式从早期到晚睡觉的因果相互作用和混响。总而言之，我们的建议旨在为重播中“噪声”结构的结构提供最强的表征事件。相关性（请参阅说明）：这项研究将为评估睡眠在大脑中的信息重组中的作用提供开放和帮助确定睡眠期间关键的时间窗口和神经元活动，这尤其重要用于信息存储和稳定。我们对性质和目的的假设和推论从医院和救济的转变期间，睡眠隐含地告知各种公共政策工人，到公立学校的早晨开始。了解睡眠的功能和机制h

项目成果

期刊论文数量（5）

专著数量（0）

科研奖励数量（0）

会议论文数量（0）

专利数量（0）

Tracing a Path for Memory in the Hippocampus.

DOI：
10.1016/j.neuron.2020.06.034
发表时间：
2020-07-22
期刊：
Neuron
影响因子：
16.2
作者：
Dutta S;Gao S;Chu JP;Kemere C
通讯作者：
Kemere C

Foraging Under Uncertainty Follows the Marginal Value Theorem with Bayesian Updating of Environment Representations.

不确定性下的觅食遵循边际值定理和环境表示的贝叶斯更新。

DOI：
10.1101/2024.03.30.587253
发表时间：
2024
期刊：
bioRxiv : the preprint server for biology
影响因子：
0
作者：
Webb,James;Steffan,Paul;Hayden,BenjaminY;Lee,Daeyeol;Kemere,Caleb;McGinley,Matthew
通讯作者：
McGinley,Matthew

Extended Poisson Gaussian-Process Latent Variable Model for Unsupervised Neural Decoding.

用于无监督神经解码的扩展泊松高斯过程潜变量模型。