Machine learning approach to identify a resting-state functional connectivity pattern serving as an endophenotype of autism spectrum disorder

Endophenotype refers to a measurable and heritable component between genetics and diagnosis, and the same endophenotype is present in both individuals with a diagnosis and their unaffected siblings. Determination of the neural correlates of an endophenotype and diagnosis is important in autism spectrum disorder (ASD). However, prior studies enrolling individuals with ASD and their unaffected siblings have generally included only one group of typically developing (TD) subjects; they have not accounted for differences between TD siblings. Thus, they could not differentiate the neural correlates for endophenotype from the clinical diagnosis. In this context, we enrolled pairs of siblings with an ASD endophenotype (individuals with ASD and their unaffected siblings) and pairs of siblings without this endophenotype (pairs of TD siblings). Using resting-state functional MRI, we first aimed to identify an endophenotype pattern consisting of multiple functional connections (FCs) then examined the neural correlates of FCs for ASD diagnosis, controlling for differences between TD siblings. Sparse logistic regression successfully classified subjects as to the endophenotype (area under the curve = 0.78, classification accuracy = 75%). Then, a bootstrapping approach controlling for differences between TD siblings revealed that an FC between the right middle temporal gyrus and right anterior cingulate cortex was substantially different between individuals with ASD and their unaffected siblings, suggesting that this FC may be a neural correlate for the diagnosis, while the other FCs represent the endophenotype. The current findings suggest that an ASD endophenotype pattern exists in FCs, and a neural correlate for ASD diagnosis is dissociable from this endophenotype. (250 words).