Identification of Synchronously Active Cells (ISAC) to study the activity of cell assemblies

Abstract

Cell assemblies--groups of neurons that fire synchronously--are central to neuronal computation. Technological advances now permit the simultaneous recording of large neuronal populations; however, many existing cell assembly detection algorithms rely on simplified measures such as pairwise correlations, failing to capture higher-order synchrony and thereby missing a rich repertoire of assemblies present in the data. To address these limitations, we developed Identification of Synchronously Active Cells (ISAC), which performs a rigorous, iterative search across all recorded units, leveraging the full dimensionality of the dataset to detect groups that fire together above chance. Using simulated recordings with ground-truth assemblies of increasing complexity, we demonstrate that ISAC considerably outperforms currently used algorithms in identifying synchronously active cells. Finally, we show that applying ISAC to experimental recordings uncovers novel features of assembly formation and dynamics. The power and sensitivity of ISAC can thus enable insights that remain inaccessible with existing methods.