Computational Classification Approach to Profile Neuron Subtypes from Brain Activity Mapping Data.
Computational Classification Approach to Profile Neuron Subtypes from Brain Activity Mapping Data.
复制标题
DOI：
10.1038/srep12474
复制
发表时间：
2015-07-27
影响因子：
4.6
通讯作者：
Tsien JZ
中科院分区：
文献类型：
--
作者：
Li M;Zhao F;Lee J;Wang D;Kuang H;Tsien JZ
行为过程中细胞类型特异性活动模式的分析对于更好地理解神经回路如何产生认知是重要的，但尚未从体内神经生理学数据集中得到很好的探索。在这里，我们描述了一种计算方法，以发现不同的细胞亚群在体内神经尖峰数据集。该方法包括四个主要步骤：放电模式特征提取、预聚类分析、聚类分类和无偏分类维度选择。通过使用两个关键特征的尖峰动态-即，伽玛分布的形状因子和尖峰间的间隔的变异系数-我们表明，这种ISICA方法提供了不变的分类多巴胺能神经元或CA 1锥体细胞亚型，无论从尖峰数据收集的大脑状态。此外，我们表明，这些ISICA分类的神经元亚型的基础不同的生理功能。我们表明，未被发现的多巴胺能神经元亚型编码不同方面的恐惧的经验，如价或价值，而不同的海马CA 1区锥体细胞差异氯胺酮诱导的麻醉反应。这种ISICA方法应该是有用的，以更好地数据挖掘的大规模在体内的神经数据集，导致新的见解与认知相关的电路动力学。
The analysis of cell type-specific activity patterns during behaviors is important for better understanding of how neural circuits generate cognition, but has not been well explored from in vivo neurophysiological datasets. Here, we describe a computational approach to uncover distinct cell subpopulations from in vivo neural spike datasets. This method, termed “inter-spike-interval classification-analysis” (ISICA), is comprised of four major steps: spike pattern feature-extraction, pre-clustering analysis, clustering classification, and unbiased classification-dimensionality selection. By using two key features of spike dynamic - namely, gamma distribution shape factors and a coefficient of variation of inter-spike interval - we show that this ISICA method provides invariant classification for dopaminergic neurons or CA1 pyramidal cell subtypes regardless of the brain states from which spike data were collected. Moreover, we show that these ISICA-classified neuron subtypes underlie distinct physiological functions. We demonstrate that the uncovered dopaminergic neuron subtypes encoded distinct aspects of fearful experiences such as valence or value, whereas distinct hippocampal CA1 pyramidal cells responded differentially to ketamine-induced anesthesia. This ISICA method should be useful to better data mining of large-scale in vivo neural datasets, leading to novel insights into circuit dynamics associated with cognitions.