Neural network embedding of functional microconnectome
Neural network embedding of functional microconnectome
复制标题
DOI：
10.1101/2021.10.19.464982
复制
发表时间：
2021-10
期刊：
影响因子：
--
通讯作者：
Arata Shirakami;T. Hase;Yuki Yamaguchi;M. Shimono
中科院分区：
文献类型：
--
作者：
Arata Shirakami;T. Hase;Yuki Yamaguchi;M. Shimono
我们的大脑就像一个复杂的网络系统。经验知识似乎被编码到有机体的网络结构中，而不是只保留个别神经元的特性。为了能够更好地考虑这种网络体系结构的高度复杂性，通过对拓扑模式的自动和可解释分析来提取简单规则将是必要的，以便允许对复杂神经体系结构内的相互关系进行更有用的观察。通过结合这两种类型的分析方法，我们可以自动压缩和自然解释功能连接性的拓扑模式，它从小鼠大脑的急性切片中同时产生电活动，持续2.5小时[Kajiwara等人。2021年]。作为第一类分析，这项研究训练了一个名为神经网络嵌入(NNE)的人工神经网络系统，并自动将功能连接性压缩到只有很小(25%)的维度。作为第二种类型的分析，我们将压缩特征与大约15个具有代表性的网络指标进行了广泛的比较，这些指标有明确的解释，包括>；5中心性类型指标和新开发的网络指标，这些指标量化了几个节点与最初关注的中心相距的程度或比率。因此，虽然我们只能对提取的特征中的55%-60%进行解释，但这些新的指标与常用的网络指标一起，能够使用自动分析对80%-100%的特征进行解释。结果不仅证明了NNE方法超越了常用的人为变量的局限性，而且也证明了认识到我们自己的局限性促使我们通过开发新的分析方法来扩展可解释的可能性的可能性。
Our brain works as a complex network system. Experiential knowledge seems to be coded into the organism’s network architecture rather than retaining only properties of individual neurons. In order to be better able to consider the high complexity of this network architecture, extracting simple rules through both automated as well as interpretable analysis of topological patterns will be necessary in order to allow more useful observations of interrelationships within the complex neural architecture. By combining these two types of analysis methods, we could automatically compress and naturally interpret topological patterns of functional connectivities, which produced electrical activities from many neurons simultaneously from acute slices of mice brain for 2.5 hours [Kajiwara et al. 2021]. As the first type of analysis, this study trained an artificial neural network system called Neural Network Embedding (NNE), and automatically compressed the functional connectivities into only small (25%) dimensions. As the second type of analysis, we widely compared the compressed features with ~15 representative network metrics, having clear interpretations, including > 5 centrality-type metrics and newly developed network metrics, that quantify degrees or ratio of hubs distanced by several-nodes from initially focused hubs. As the result, although we could give interpretations for only 55-60% of the extracted features, these new metrics, together with the commonly utilized network metrics, enabled interpretations for 80-100% features, using automated analysis. The result demonstrates not only the fact that the NNE method surpasses limitations of commonly used human-made variables, but also the possibility that acknowledgement of our own limitations drives us to extend interpretable possibilities by developing new analytic methodologies.