Feature overlap in models of sustained attention and autism symptoms

Whole-brain edge-behavior patterns in connectome-based predictive models are reproducible within ADHD/autism or sustained-attention phenotypes (r > 0.13, P < 0.005) but largely distinct across phenotypes.

• Why it matters: It is unclear whether CPM features for sustained attention and autism symptoms reflect shared neurobiology or phenotype-specific mechanisms, despite cross-sample generalization.
• What they did: The study compared feature overlap across 5 published CPMs (top 1,000 positively and negatively predictive edges per model) using within-phenotype consistency tests and cross-phenotype specificity, plus representational similarity analysis.
• The result: No edge or network pair was shared by all 5 models; only 1 of 6 cross-phenotype pairs correlated significantly (negatively), and autism-vs-attention models shared no more edges than chance. Scan type, participant age, and autism diagnosis did not explain similarity across datasets.

AI summary of the abstract · may contain errors · not medical advice · read the paper.