A structural vulnerability model of autism derived from symptom-ability network analysis

While Autism Spectrum Disorder (ASD) is increasingly understood through a developmental systems lens, the structural properties that confer vulnerability remain poorly defined. This study aimed to formalize and test a "Structural Vulnerability Model" by comparing the symptom-ability network architectures of preschool children with and without ASD.

Utilizing a large dataset (N = 1,502 ASD; N = 570 Typically Developing [TD]), we constructed Gaussian Graphical Models and Directed Acyclic Graphs (DAGs) from ten variables spanning core symptoms, adaptive behaviors, and cognitive abilities. We performed a multi-level comparison of network topology, node centrality, gender-specific patterns, and conditional-dependence structures.

The ASD and TD networks differed in global strength and in the pattern of prominent edges. Communication (COM) had the highest strength centrality in TD, whereas Receptive Language (RL) had the highest strength in ASD. Exploratory bridge-centrality analyses identified RL and COM as prominent cross-domain connectors in ASD. The cross-sectional DAG analysis summarized conditional-dependence structures and was not interpreted causally.

The findings provide preliminary, network-based support for a Structural Vulnerability Model, interpreted as a pattern of concentrated associations rather than a validated causal or prognostic mechanism. Longitudinal, externally replicated, and covariate-adjusted studies are needed before intervention implications can be drawn.