Abstract

Connected networks are central to neurobiology, and reconstructing them is key to understanding the computational principles of the brain. Modern recording technologies—calcium imaging, Neuropixels, and large-scale electrophysiology—now generate massive neuronal datasets, creating a need for robust, interpretable graph-based analysis tools. We present a simple graph reconstruction framework [1] enhanced by a data-driven adaptive thresholding method derived from statistics. Using strongly connected components, we extract functionally relevant sub-networks, and we characterize their topology through Markovian statistics [2,3]. Applied to volumetric calcium imaging data [4], our approach reveals sub-networks whose activity patterns suggest their role as fundamental computational units of the brain. [1] Aymard, P., Boffi J-C., Asari H., Prevedel R., Holcman D. Column-Like Subnetwork Reconstruction in Motor Cortex  from Graph-Based 3D High-Density Two-Photon Calcium Imaging doi: https://doi.org/10.1101/2025.06.17.660119. [2] Karlin, S., and Taylor, H. (1981). A Second Course in Stochastic Processes. vol. 2. Elsevier Science.[3] Boyd, S., Diaconis, P., and Xiao, L. (2004). Fastest mixing markov chain on a graph. SIAM Review 46.[4] Prevedel, R., Verhoef, A., Pernía-Andrade, A. et al. Fast volumetric calcium imaging across multiple cortical layers using sculpted light. Nat Methods 13.