Correlated Gene Expression Supports Neuromorphic Epicenter Identification in Attention-Deficit/Hyper

388 

Abstract Submission 

Pan Wang¹, jinzhong peng¹, Qingquan Cao¹, Yilu Li¹, Bharat Biswal²

¹UESTC, Chengdu, Sichuan, ²New Jersey Institute of Technology, Newark, NJ

Pan Wang  
UESTC
Chengdu, Sichuan

jinzhong peng  
UESTC
Chengdu, Sichuan

Qingquan Cao  
UESTC
Chengdu, Sichuan

Yilu Li  
UESTC
Chengdu, Sichuan

Bharat Biswal  
New Jersey Institute of Technology
Newark, NJ

Attention-deficit/hyperactivity disorder (ADHD), a highly heritable developmental psychiatric disorder, primarily manifested the inattentive, hyperactive, and impulsive symptoms (Gallo and Posner, 2016; Lord et al., 2018; Posner et al., 2020). The neuromorphic heterogeneity and how the correlated gene expression (CGE) connectome influences morphological change in ADHD have not been investigated.

Current study employed the neuroimaging dataset from a publicly available ADHD-200 dataset (http://fcon_1000.projects.nitrc.org/indi/adhd200/), including 196 ADHD patients and 181 healthy controls (HCs). Adopting the neuroimaging subtype analysis based on W-score, we estimated the cortical thickness deviation, and further obtained the biotypes of ADHD using the density peak-based clustering analysis (Figure 1A). To construct the correlated gene expression (CGE) connectome matrix, we first obtained the regional matrix of transcriptional level (400 regions × 15,631 gene expression) in line with previous study on CGE (Arnatkeviciute et al., 2019), and then calculated the transcriptomic similarities between the distinct regional gene expressions, resulting in a symmetric CGE connectome matrix (400× 400) (Figure 1B). Partial least squares (PLS) gene list based on W-score map and WCGE-score map were put into Metascape for gene enrichment analysis (Figure 1C). Cortical epicenter regions were identified by putative epicenters if their deviation was high and their neighbors also experienced high deviation (Figure 1D).

Findings from clustering analysis revealed that ADHD patients could be divided into two discriminative biotypes. The significant regions in biotype 1 primarily located in the visual peripheral network, control network and default mode network, while regions for biotype 2 mainly distributed in the salience ventral attention network. The CGE connectome exhibited modular organization with distinct robust pattern similar as previous study (Romero-Garcia et al., 2018; van den Heuvel et al., 2019). We observed that the regional W-score values were significantly positive correlations with the CGE-informed W-score for both biotypes (Biotype 1: R = 0.4574, P < 0.0001; biotype2: R = 0.2720, P < 0.0001). In addition, patterns of epicenter likelihood revealed that biotype 1 was mainly associated with the lateral prefrontal lobe and temporo-parietal junction, while biotype 2 were relating to the lateral sulcus and medial prefrontal lobe (Figure 2A and 2B). More importantly, we found 6 and 8 overlapped regions between epicenters and the top 5% regions from W-score in biotype 1 and biotype 2, respectively (Figure 2C).

The present study demonstrated that ADHD patients could be classified into two biotypes closely relating to CGE connectome by estimating the neighborhood cortical thickness based on regional W-score measures. Epicenter identification would promote understanding of the heterogeneity and distinct clinical manifestations in ADHD biotypes.

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

Transcriptomics ²

FUNCTIONAL MRI

Neurological

Psychiatric Disorders

Structures