Normative models of individualized functional brain networks reveal language network expansion in autism

Abstract

Autism spectrum disorder is a highly heterogeneous neurodevelopmental disorder, hindering mechanistic insights and the identification of biomarkers for clinical diagnosis. Recently, precision functional mapping has been developed to identify abnormalities in brain network topologies associated with various psychiatric disorders, yet its application in autism remains limited. Here, we utilized precision functional mapping and a large, multisite neuroimaging dataset (N = 1,182) to construct individualized functional networks in individuals with autism. We developed normative models using network surface area from healthy controls (n = 628) to characterize typical brain network organization across age, allowing for the quantification of individual-specific deviations in individuals with autism (n = 554). We found widespread and heterogeneous deviations from the normative model, with the language network emerging as the most significantly altered region, thereby emerging as an epicenter of functional disruption in autism. Individuals with autism were clustered into three subtypes involving distinct functional network topologies, associated with behavioral profiles marked by impairments in perception, language processing, or socio-emotional functioning. We further linked these atypical brain features to cortical gene expression patterns, revealing enriched pathways related to neurodevelopment, language, and signaling processes. Together, these findings reveal autism-specific deviations in individualized functional brain networks, offering potential clinical relevance for understanding and stratifying autism.