Identification of sex differences in autism using class imbalance mitigated functional connectivity

364 

Abstract Submission 

Jong Young Namgung¹, Jong Min Mun², YEONGJUN PARK³, Jaeoh Kim^1,4, Bo-yong Park^1,4,5

¹Department of Data Science, Inha University, Incheon, Republic of Korea, ²Marshall School of Business, University of Southern California, Los Angeles, CA, United States, ³Department of Computer Engineering, Inha University, Incheon, Republic of Korea, ⁴Department of Statistics and Data Science, Inha University, Incheon, Republic of Korea, ⁵Center for Neuroscience Imaging Research, Institute for Basic Science, Suwon, Republic of Korea

Jong Young Namgung  
Department of Data Science, Inha University
Incheon, Republic of Korea

Jong Min Mun  
Marshall School of Business, University of Southern California
Los Angeles, CA, United States

YEONGJUN PARK  
Department of Computer Engineering, Inha University
Incheon, Republic of Korea

Jaeoh Kim  
Department of Data Science, Inha University|Department of Statistics and Data Science, Inha University
Incheon, Republic of Korea|Incheon, Republic of Korea

Bo-yong Park  
Department of Data Science, Inha University|Department of Statistics and Data Science, Inha University|Center for Neuroscience Imaging Research, Institute for Basic Science
Incheon, Republic of Korea|Incheon, Republic of Korea|Suwon, Republic of Korea

Autism spectrum disorder (ASD) is a common psychiatric condition during development, and individuals with ASD show impaired social interaction skills and restricted/repetitive behaviors (Mottron et al. 2006). The natural characteristic of ASD is that males tend to be diagnosed more frequently than females (Werling and Geschwind 2013). Due to the sex imbalance in ASD, we lack an understanding of the differences in connectome organization of the brain between male and female individuals with ASD. In this study, we matched the sex ratio using a Gaussian mixture model-based oversampling technique and investigated the differences in functional connectivity between male and female ASDs.

We obtained T1-weighted structural MRI and resting-state functional MRI (rs-fMRI) data of 507 individuals with ASD (mean ± SD age = 17.08 ± 8.50 years; 12.22 % female) and 553 typically developing controls (mean ± SD age = 17.10 ± 7.74 years; 17.35 % female) from Autism Brain Imaging Data Exchange I initiative (ABIDE-I) (Di Martino et al. 2014), and the imaging data was preprocessed using micapipe (Cruces et al. 2022). The functional connectivity matrix was constructed by calculating linear correlations of time series between different brain regions defined using the Schaefer atlas with 300 parcels (Schaefer et al. 2018). We applied diffusion map embedding to estimate low-dimensional representations of functional connectivity (i.e., gradients) (Margulies et al. 2016; Vos de Wael et al. 2020; Coifman and Lafon 2006) and summarized it according to seven intrinsic functional networks (Thomas Yeo et al. 2011). To adjust the class imbalance problem between sexes, we estimated the gradient distribution of the female group using the Gaussian mixture model and generated synthetic samples from the estimated distributions. Then, we assessed the interaction effects of sex and group to evaluate sex-related differences in functional gradients. The significance was assessed using 1,000 permutation tests, and multiple comparisons across brain networks were corrected using a false discovery rate (FDR) < 0.05 (Benjamini and Hochberg 1995).

The generated functional gradient differentiated sensory regions and default mode areas (Fig. 1a). The oversampled gradients of females with ASD and control groups showed similar spatial patterns with the actual data. When we assessed the interaction effect between sex and group, the default mode network showed a significant effect (t = -5.63, p-perm-FDR < 0.001; Fig. 1b). In addition, the visual (t = 2.22, p-perm-FDR = 0.098), somatomotor (t = 2.63, p-perm-FDR = 0.079), and dorsal attention networks (t = 2.73, p-perm-FDR = 0.079) showed moderate effects. Specifically, the gradient values decreased in females with ASD than controls in the default mode network, while those of male ASDs did not change. The gradient values of the sensory and attention networks showed opposite patterns.

In this study, we opted for the Gaussian mixture model-based oversampling approach to mitigate sex imbalance in the ASD dataset and observed significant sex-related differences in functional gradients in individuals with ASD. Our systematic analyses may provide insights for understanding the heterogeneity of ASD.

Funding:
National Research Foundation of Korea (NRF-2021R1F1A1052303; NRF-2022R1A5A7033499), Institute for Information and Communications Technology Planning and Evaluation (IITP) funded by the Korea Government (MSIT) (No. 2022-0-00448, Deep Total Recall: Continual Learning for Human-Like Recall of Artificial Neural Networks; No. RS-2022-00155915, Artificial Intelligence Convergence Innovation Human Resources Development (Inha University); No. 2021-0-02068, Artificial Intelligence Innovation Hub), Institute for Basic Science (IBS-R015-D1).

Neurodevelopmental/ Early Life (eg. ADHD, autism) ¹

Bayesian Modeling

Connectivity (eg. functional, effective, structural)

fMRI Connectivity and Network Modeling ²

BOLD fMRI

Autism

Cortex

Data analysis

FUNCTIONAL MRI

Sexual Dimorphism