Aberrant multivariate mapping between behavioral profiles and cortical morphological brain networks in children with autism spectrum disorder

Understanding how human behavior relates to brain structure and function is a central goal of neuroscience. Autism spectrum disorder (ASD) is increasingly conceptualized as a disorder of connectome dysfunction; however, how morphological brain networks relate to behavioral profiles in children with ASD remains unclear. In this study, we collected a comprehensive battery of behavioral assessments and structural MRI data from 101 children with ASD and 67 typically developing (TD) children. Individual morphological brain networks were constructed by integrating four cortical features to estimate morphological connectivity (MC). We then applied sparse canonical correlation analysis (sCCA) to identify multivariate brain-behavior associations within each group. Using group-specific MC features, four distinct association modes were identified in TD children, none of which generalized to the ASD group. In ASD children, three modes were detected that captured comparable behavioral dimensions with three TD modes but were supported by largely distinct neural architectures. Within a unified MC framework, an autistic trait mode emerged in both groups; however, it was primarily supported by MC between the default and control networks in TD children, versus MC between the default and dorsal attention networks in ASD children. Notably, the prosociality mode present in TD children was consistently absent in the ASD group, regardless of the MC features used in sCCA. These findings reveal a reorganization of brain-behavior relationships in children with ASD and provide evidence for a neurodiverse neural architecture, in which similar behaviors arise from divergent neurodevelopmental pathways, moving beyond a simple continuum model of typical development.