Personalized circuit modeling captures variation in cortical functional connectivity

Abstract

Functional magnetic resonance imaging (fMRI) of the human cortex reveals patterns of correlated neural dynamics that are individual-specific and associated with phenotypic variation. However, circuit mechanisms underlying individual variation in functional connectivity (FC) are not well understood. Here, we fit individual-level FC patterns with a biophysically-based circuit model of large-scale cortical dynamics. This model is fit with a small number of neurophysiologically interpretable parameters, and incorporates a hierarchical gradient in local synaptic strengths across cortex parameterized via the structural MRI-derived T1w/T2w map. We applied our modeling framework to resting-state fMRI FC from a large cohort of subjects (N=842) from the Human Connectome Project. We found that the model captures a substantial portion of individual variation in FC, especially with personalized degrees of local synaptic specialization along the hierarchical gradient. Furthermore, the model can capture to the within-subject variation in FC across scans. Empirically, we found that principal modes of individual variation in FC follow interpretable topographic patterns. We developed a framework to assess model expressivity via how these empirical modes of FC variation align with variations in simulated FC induced by parameter perturbations. This framework reveals a straightforward mapping between key parameters and the leading modes of variation across subjects and provides a principled approach to extending computational models. Collectively, our modeling results establish a foundation for personalized computational modeling of functional dynamics in large-scale brain circuits.