Functional connectivity-based subtypes of individuals with and without autism spectrum disorder

Abstract

Autism spectrum disorder (ASD) is a heterogeneous neurodevelopmental disorder, characterized by impairments in social communication and restricted, repetitive behaviors. Neuroimaging studies have shown complex patterns and functional connectivity (FC) in ASD, with no clear consensus on brain-behavior relationships or shared patterns of FC with typically developing controls. Here, we used a dimensional approach to characterize two distinct clusters of FC patterns across both ASD participants and controls using k-means clustering. Using multivariate statistical analyses, a categorical approach was taken to characterize differences in FC between subtypes and between diagnostic groups. One subtype was defined by increased FC within resting-state networks and decreased FC across networks compared with the other subtype. A separate FC pattern distinguished ASD from controls, particularly within default mode, cingulo-opercular, sensorimotor, and occipital networks. There was no significant interaction between subtypes and diagnostic groups. Finally, a dimensional analysis of FC patterns with behavioral measures of IQ, social responsiveness, and ASD severity showed unique brain-behavior relations in each subtype and a continuum of brain-behavior relations from ASD to controls within one subtype. These results demonstrate that distinct clusters of FC patterns exist across ASD and controls, and that FC subtypes can reveal unique information about brain-behavior relationships.

Keywords: Autism spectrum disorder; Brain-behavior relationships; Clustering; Functional connectivity; Multivariate statistics; Resting-state networks.

Figure 1.

(A) Elbow point plots, indicating that the optimal number of clusters is 2. (B) Bootstrapping procedure to evaluate the reliability of the optimal number of clusters. Error bars show 95% confidence intervals. (C) Subtype centroids. DMN = default mode network; FPN = fronto-parietal network; CON = cingulo-opercular network; ON = occipital network; SMN = sensorimotor network; CN = cerebellar network.

Figure 2.

Results from the multivariate group analysis. (A) First pattern and (B) second pattern, and the associated BSRs for each connection at a threshold of ±2. Error bars show 95% confidence intervals determined through bootstrap resampling.

Figure 3.

Significant contributions of RSN pairs to each pattern for positive and negative BSRs, for the (A) first pattern and (B) second pattern from the multivariate group analysis. Orange = positive BSRs, green = negative BSRs.

Figure 4.

Results from the multivariate brain-behavior analysis. (A) First pattern and (B) third pattern, and the associated BSRs for each connection at a threshold of ±2. Error bars show 95% confidence intervals determined through bootstrap resampling.

Figure 5.

Significant contributions of RSN pairs to each pattern for positive and negative BSRs for 9A) first pattern and (B) third pattern. Orange = positive BSRs, green = negative BSRs.

Figure 6.

Brain and behavior scores for Subtype 2, from the first pattern of the multivariate brain-behavior analysis.