Neural biomarkers distinguish severe versus mild autism spectrum disorder amongst high-functioning individuals

Abstract

Background: While increasing evidence in neuroscience has been advancing our understanding of Autism Spectrum Disorder (ASD), the relatively small effect sizes prevent us from pinning down any neural biomarkers for diagnostic purposes. There is therefore a need for identifying stratification biomarkers to parse this diverse condition into more homogeneous subgroups, such as mild versus severe ASD. Methods: Study samples (ASD group, n=260; Control group, n=574) were derived from ABIDE I and an independent validation sample from ABIDE II (v-ASD group, n=29). Canonical correlation analysis and hierarchical clustering were used to partition ASD group into subgroups. Support vector machine (SVM) were trained through the leave-one-out strategy to predict individual’s ADOS score within the ASD group, which was further validated in v-ASD group. Results: The FC-based partition derived two subgroups which represented severe (n=169) versus mild (n=91) ASD patients. The SVM model found moderate fitness with the clinically rated ADOS total score in the ASD group (r=0.24, pone-tailed<0.0001), and was successfully validated in v-ASD group (r=0.32, ppermutation=0.0385). FCs between temporal areas, amygdala, anterior cingulate cortex, postcentral gyrus and left inferior frontal gyrus were found with graded changes in strength from controls, ASD-mild, to ASD-severe groups, whereas FCs between prefrontal areas, thalamus and postcentral gyrus were found specific to the mild ASD group.Discussion: The current study provided multiple pieces of evidence with replication to show that rsfMRI FCs can serve as candidate stratification neural biomarkers in partitioning high-functioning autistic individuals based on their symptom severity. Clinical relevance were also discussed.

More information

Identifiers

ORCID for Yuning Zhang: orcid.org/0000-0003-2225-6368

Catalogue record

Export record

Altmetrics