Department of Mathematics and Statistics, University of Turku, Turku, Finland.
Université Grenoble Alpes, CEA, INSERM, Biology of Cancer and Infection UMR S 1036, Grenoble, France.
Bioinformatics. 2018 Sep 15;34(18):3196-3204. doi: 10.1093/bioinformatics/bty322.
The collective and co-ordinated synaptic activity of large neuronal populations is relevant to neuronal development as well as a range of neurological diseases. Quantification of synaptically-mediated neuronal signalling permits further downstream analysis as well as potential application in target validation and in vitro screening assays. Our aim is to develop a phenotypic quantification for neuronal activity imaging data of large populations of neurons, in particular relating to the spatial component of the activity.
We extend the use of Markov random field (MRF) models to achieve this aim. In particular, we consider Bayesian posterior densities of model parameters in Gaussian MRFs to directly model changes in calcium fluorescence intensity rather than using spike trains. The basis of our model is defining neuron 'neighbours' by the relative spatial positions of the neuronal somata as obtained from the image data whereas previously this has been limited to defining an artificial square grid across the field of view and spike binning. We demonstrate that our spatial phenotypic quantification is applicable for both in vitro and in vivo data consisting of thousands of neurons over hundreds of time points. We show how our approach provides insight beyond that attained by conventional spike counting and discuss how it could be used to facilitate screening assays for modifiers of disease-associated defects of communication between cells.
We supply the MATLAB code and data to obtain all of the results in the paper.
Supplementary data are available at Bioinformatics online.
大量神经元群体的集体和协调的突触活动与神经元发育以及一系列神经疾病有关。对突触介导的神经元信号进行量化可以进行进一步的下游分析，并有可能应用于靶标验证和体外筛选测定。我们的目标是为大量神经元群体的神经元活动成像数据开发一种表型量化方法，特别是与活动的空间组成部分有关。
我们扩展了马尔可夫随机场 (MRF) 模型的使用来实现这一目标。具体来说，我们考虑了高斯 MRF 中模型参数的贝叶斯后验密度，以直接对钙荧光强度的变化进行建模，而不是使用尖峰列车。我们模型的基础是通过从图像数据中获得的神经元体的相对空间位置来定义神经元“邻居”，而以前这仅限于在视场中定义一个人工正方形网格和尖峰分箱。我们证明了我们的空间表型量化方法适用于包含数百个时间点的数千个神经元的体外和体内数据。我们展示了我们的方法如何提供超越传统尖峰计数获得的见解，并讨论了它如何用于促进用于修饰细胞间通信的疾病相关缺陷的筛选测定。
我们提供了 MATLAB 代码和数据，以获得本文中的所有结果。
补充数据可在 Bioinformatics 在线获取。
