Major depressive disorder (MDD) is a prevalent mental illness, with severe symptoms that can significantly impair daily routines, social interactions, and professional pursuits. Recently, imaging genetics has received considerable attention for understanding the pathogenesis of human brain disorders. However, identifying and discovering the imaging genetic patterns between genetic variations, such as single nucleotide polymorphisms (SNPs), and brain imaging data still present an arduous challenge. Most of the existing MDD research focuses on single-modality brain imaging data and neglects the complex structure of brain imaging data.
In this study, we present a novel association analysis model based on a self-expressive network to identify and discover imaging genetics patterns between SNPs and multi-modality imaging data. Specifically, we first build the multi-modality phenotype network, which comprises voxel node features and connectivity edge features from structural magnetic resonance imaging (sMRI) and resting-state functional magnetic resonance imaging (rs-fMRI), respectively. Then, we apply intra-class similarity information to construct self-expressive networks of multi-modality phenotype features via sparse representation. Subsequently, we design a fusion method guided by diagnosis information, which iteratively fuses the self-expressive networks of multi-modality phenotype features into a single new network. Finally, we propose an association analysis between MDD risk SNPs and the multi-modality phenotype network based on a fusion self-expressive network.
Experimental results show that our method not only enhances the association between MDD risk SNP rs1799913 and the multi-modality phenotype network but also identifies some consistent and stable regions of interest (ROIs) multi-modality biological markers to guide the interpretation of MDD pathogenesis. Moreover, 15 new potential risk SNPs highly associated with MDD are discovered, which can further help interpret the MDD genetic mechanism.
In this study, we discussed the discriminant and convergence performance of the fusion self-expressive network, parameters, and atlas selection.