Identifying Diagnostic Biomarkers for Autism Spectrum Disorder From Higher-order Interactions Using the PED Algorithm

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Developmental Neuroscience

Background:

• Autism spectrum disorder (ASD) research often overlooks higher-order brain region interactions.
• Existing neuroimaging studies primarily focus on pairwise brain region connectivity.

Purpose of the Study:

• To explore complex, higher-order interactions among brain regions in individuals with ASD.
• To develop and validate a method for identifying key brain triads and their diagnostic potential in ASD.

Main Methods:

• Utilized partial entropy decomposition (PED) to compute higher-order dependencies within brain triads.
• Employed surrogate tests to assess the influence of individual brain regions on triads.
• Applied hypergraph modularity maximization to reveal higher-order brain structures.
• Developed a classification model to evaluate the diagnostic utility of identified key triads.

Main Results:

• Identified specific key triads with altered interactions in ASD compared to typical controls (TC).
• Observed a looser connection between the right and left thalamus in ASD.
• Found attenuated interaction in a redundant key triad (left cerebellum crus 1, left precuneus, right inferior occipital gyrus) in ASD.
• Detected a notable decline in a synergistic key triad (right cerebellum crus 1, left postcentral gyrus, left lingual gyrus) in ASD.

Conclusions:

• The proposed method effectively captures higher-order brain interactions relevant to ASD.
• Specific key triads demonstrate potential as reliable diagnostic biomarkers for autism spectrum disorder.
• Findings highlight the importance of examining higher-order interactions in neurodevelopmental disorders.