An engine for systematic discovery of cause-effect relationships between brain structure and function

Abstract

Characterising how perturbations of brain architecture influence brain function is essential to understand the origins of brain dysfunction, and devise potential avenues of treatment. Here we introduce a computational engine for systematic causal discovery of the functional consequences of altering network architecture and local biophysics in the brain. We integrate multimodal anatomical and functional neuroimaging to implement over 2, 000 in-silico brains, and provide mechanistic insight into the functional consequences of local lesions, global wiring, and empirically-derived maps of regional cytoarchitecture and chemoarchitecture. We comprehensively assess how each manipulation of brain macrostructure reshapes spatial and temporal signal coordination, information dynamics, and functional hierarchy--as well as spontaneous co-activation of meta-analytic cognitive circuits, and > 6, 000 dimensions of local neural dynamics. Our computational model systematically identifies which features of brain architecture have overlapping or antagonistic causal influence over each dimension of brain function, and how functional properties are traded off against each other across disorders and neuromodulation. We find that regions functional vulnerability to lesions in silico recapitulates their vulnerability to neurodevelopmental and psychiatric in vivo, along a core-periphery organisation. We provide convergent evidence that the brains wiring diagram is finely tuned to favour the hierarchical integration of information. Notably, our model successfully recapitulates known empirical results that have not been modelled before, including desynchronisation and flattening of the brains functional hierarchy induced by psychedelic 5HT2A agonists. To catalyse future discoveries, we make this resource freely available to the neuroscience community through an interactive website (https://systematic-causal-mapping.up.railway.app/), where users can interrogate our systematic database of simulated cause-effect relationships. Altogether, we provide a powerful computational engine to predict the functional consequences of experimental or clinical interventions, and drive neuroscientific hypothesis-generation.