Beyond pairwise interactions: Exploring brain high-order interactions in neurodegeneration and states of consciousness

Understanding the brain requires recognizing interactions beyond simple pairs of regions. Recent advances in multivariate information theory enable measuring high-order interactions from neural recordings. These metrics assess the level of interdependence among groups of three or more variables and the dominant quality of the interactions, such as synergy or redundancy, revealing complex patterns in neural dynamics. In this work, I will demonstrate their application in various contexts. First, I will show that brain high-order interactions are significantly disrupted in neurodegeneration, with a marked tendency towards high-order hypoconnectivity. Second, by analyzing sleep EEG data, I will demonstrate that these metrics can predict the risk of developing dementia. Third, I will show that they can track changes in consciousness induced by mind-altering drugs. Finally, I will discuss the current challenges of the high-order approach in neuroscience and propose a spectrum of strategies based on interaction order and levels of observation.