ƽÌØÎå²»ÖÐ

Graph and geometry-based approaches for cellular-level analysis of structure and function in neuronal networks

Patrick Desrosiers, Université Laval/CERVO
Tuesday November 26, 12-1pm
Zoom Link:Ìý
In Person: 550 Sherbrooke, Room 189

Abstract: Network science, primarily inspired by graph theory and computer science, has deepened our understanding of brain circuits. However, this progress is based predominantly on studies of the human brain using MRI techniques, which do not capture detailed cellular information. In this talk, I will show that graph-theoretical methods can successfully be applied at the cellular level in different imaging contexts, ranging from quantitative phase imaging of neuronal cultures using digital holographic microscopy (DHM) to whole-brain calcium imaging in living zebrafish larvae. In rat neuron cultures, improvements in cell segmentation for DHM data and the development of a graph inference method enable accurate classification of dozens of neuronal cultures into developmental stages. When applied to hiPSC-derived neuron cultures from schizophrenia and control cohorts, this approach reveals significant functional differences between the cohorts, though not accompanied by structural ones, aligning with the hypothesis of NMDA receptor hypofunctionality in schizophrenia. Moreover, our work on zebrafish larvae demonstrates that mesoscopic functional connectivity is a reproducible measure of brain activity that reflects individuality. Using thousands of single-neuron reconstructions, we find strong coupling between functional and structural connectivity, as well as functional network gradients that map onto the sensorimotor functions of brain regions. Lastly, revisiting recent claims about geometric constraints on brain function, we explore the correspondence between geometric eigenmodes and functional gradients of neuronal activity within three-dimensional structures through numerical simulations and experimental observations in larval zebrafish. This recent work underscores the critical role of short-range connections in the emergence of geometric features in neuronal activity and cautions against misinterpretations in filtered neuroimaging data at larger observational scales.

Short bio: Initially trained in physics and mathematics at Université Laval, the University of Melbourne, and CEA-Saclay, Patrick Desrosiers is a neuroscience researcher at the CERVO Brain Research Center and an adjunct professor of physics at Université Laval. As co-director of Dynamica, a multidisciplinary research group in complex systems, he develops theoretical and computational methods to better understand the relationship between structure and function in neural networks, with a focus on dimensionality reduction and network resilience to perturbations.