Uncovering functional connectivity patterns predictive of cognition in youth using interpretable predictive modeling.

Brain-wide association studies using functional MRI have advanced our understanding of how behavioral traits relate to individual variability in brain function. These studies typically identify functional connectivity (FC) patterns linked to behavioral traits using either whole-brain or region-wise predictive models. However, whole-brain models often struggle with generalizability and interpretability due to the high dimensionality of FC data, while region-wise models isolate predictions, limiting their ability to capture the integrated contributions of brain-wide FC patterns. In this study, we introduce an interpretable predictive model that learns fine-grained FC patterns predictive of behavioral traits, jointly at the regional and participant levels, to characterize the overall association of FC patterns with a target trait. Our model jointly learns a relevance score and a dedicated prediction function for each brain region, then integrates the regional predictions using the relevance scores as weights to generate a participant-level prediction, capturing the collective association of FC patterns with the trait. We validated our method using FC data from 6,798 participants in the Adolescent Brain and Cognitive Development (ABCD) study to predict cognition. Our model identified the cingulo-parietal, retrosplenial-temporal, dorsal attention, and cingulo-opercular networks as collectively predictive of cognitive traits, achieved competitive prediction accuracy, and enabled detailed characterization of fine-grained FC differences across cognitive domains. The learned relevance scores enhanced region-wise predictions of longitudinal cognitive measures in the ABCD cohort and cognitive traits in the Human Connectome Project Development cohort. These findings suggest that our method effectively characterizes generalizable and fine-grained FC patterns linked to cognition in youth.