Application of deep learning algorithm on whole genome sequencing data uncovers structural variants associated with multiple mental disorders in African American patients

This study investigated the association between genomic structural variants and eight common mental disorders in African American patients using a deep learning algorithm. Whole genome sequencing data from 4179 individuals (1384 patients with at least one mental disorder) were analyzed. The model achieved ~65% accuracy in differentiating patients from controls and showed promising results in labeling patients with multiple disorders. Genes in highly weighted genomic regions were enriched in pathways related to immune responses and G-protein receptor activities. Variants in non-coding regions performed comparably to coding region variants but lacked genomic hotspots, suggesting they may serve as alternative markers.

Yichuan Liu, Hui-Qi Qu, Frank D. Mentch, Jingchun Qu, Xiao Chang, Kenny Nguyen, Lifeng Tian, Joseph Glessner, Patrick M. A. Sleiman, Hakon Hakonarson