CRCNS US-French Research Proposal: Architectural Principles and Predictive Modeling of the Mammalian Connectome

University Of Notre Dame, Notre Dame IN

Investigators

Abstract

This US-France collaborative project is aimed at discovering the fundamental properties of the structural/anatomical organization of cortical connections capable of supporting massive amounts of computations in the brain, despite the 100,000-fold variation in mass from the smallest mammals to the largest. Several independent empirical observations (such as sensory substitution experiments) suggest the existence of common network architectural principles in the mammalian cortex, critical for efficient and hierarchically modular information processing. Through capturing these fundamental structural and dynamical features in large-scale neuronal networks across several species, this project will help with our understanding of information processing in the human brain. It will also inform the emerging field of neuromorphic engineering, which focuses on bio-inspired computational devices. The outcomes of this project may also be relevant to neuro-degenerative diseases, given the growing evidence suggesting that disease progression often occurs via the breakdown of high-centrality, long-range connections between cortical areas, which will be characterized within this project. By extending empirical, consistent tract-tracing databases for the physical network of interareal cortical connections in the macaque and mouse (supplemented by dMRI tractography data) and exploiting recent discoveries related to the Exponential Distance Rule (EDR) (which has been empirically demonstrated in several mammals), this project aims to capture the network architectural invariants of the cortex. These invariants are graph theoretical properties of the connectome that are preserved across mammalian brains and across scales. Based on recent empirical evidence, the project puts forward the hypothesis that the EDR also plays a critical role in generating sparse encoding of highly correlated information streams in a scale-invariant manner, a hypothesis that will be tested within a predictive modeling approach. The work will also generate novel imputation algorithms suitable for dense networks and novel, efficient algorithms for comparing species connectomes, exploiting the spatial embeddedness of these networks. A companion project is being funded by the French National Research Agency (ANR).