Detecting cell assembly interaction patterns via Bayesian based change-point detection and graph inference model

Abstract

Recent studies have proposed the theory of functional network-level neural cell assemblies and their hierarchical organization architecture. In this study, we first proposed a novel Bayesian binary connectivity change point model to be applied on the binary spiking time series recorded from multiple neurons in the mouse hippocampus during three different emotional events, to find stable temporal segments of neural activity. We then applied a Bayesian graph inference algorithm on the segmentation results to find multiple functional interaction patterns underlying each experience. The resulting interaction patterns were analyzed by multi-view co-training method to identify the common sub-network structure of cell assemblies which are strongly connected i.e. "neural cliques". By analyzing the resulting sub-networks from three memory-producing events, it is found that there exist certain common neurons participating in the functional interactions across different events, lending strong support evidence to the hypothesis of hierarchical organization architecture of neuronal assemblies.