Predicting individual differences in social functioning from functional connectomes in children with autism spectrum disorder.

Autism spectrum disorder (ASD) is characterized by profound heterogeneity in social functioning. While previous studies using group-level comparisons and predefined regions of interest have identified neural correlates of social deficits in ASD, the neural basis of individual differences in social functioning remains unclear. Here, we applied connectome-based predictive modeling to resting-state functional MRI data from 65 children with ASD aged 1.5-8 years (mean age = 4.23 years) to predict individual differences in social functioning assessed by the personal-social subscale of the Griffiths Development Scales (GDS). Results showed that whole-brain functional connectivity significantly predicted personal-social scores. The positive predictive network was dominated by connections within the default mode network, with the left precuneus emerging as a core hub. In contrast, the negative predictive network mainly involved inter-network connections, particularly between the frontoparietal and sensorimotor networks. In an independent cohort of 27 children with ASD aged 1.25-8 years (mean age = 3.31 years), the positive-edge model showed a significant association with personal-social scores, whereas the negative-edge and combined models did not. These findings suggest that intrinsic large-scale brain network organization is associated with individual differences in social functioning in children with ASD, highlighting the potential of network-level approaches for characterizing neurodevelopmental heterogeneity and supporting future precision-oriented research.