Integration of Imaging Genomics Data for the Study of Alzheimer's Disease Using Joint-Connectivity-Based Sparse...

Area of Science:

• Neuroimaging
• Genetics
• Biomarker Discovery

Background:

• Imaging genetics links brain structure and genetic variations.
• Identifying robust biomarkers for Alzheimer's disease (AD) remains challenging.
• Prior knowledge integration is crucial for biologically meaningful discoveries.

Purpose of the Study:

• To develop a novel algorithm for integrating multimodal data (sMRI, SNPs, gene expression) to identify AD biomarkers.
• To leverage brain connectivity and genetic information as prior knowledge for enhanced biomarker discovery.
• To improve the robustness and interpretability of imaging genetics analyses.

Main Methods:

• Proposed joint-connectivity-based sparse nonnegative matrix factorization (JCB-SNMF).
• Simultaneously projected structural MRI, SNP, and gene expression data into a common feature space.
• Incorporated brain connectivity and genetic data as prior knowledge, using GraphNet regularization.

Main Results:

• JCB-SNMF demonstrated superior anti-noise performance compared to existing NMF methods.
• Identified SF3B1, RPS20, and RBM14 as potential AD biomarkers via PPI network analysis.
• Discovered significant SNP-ROI and gene-ROI associations, including specific SNPs and genes potentially affecting gray matter volume.

Conclusions:

• JCB-SNMF offers a powerful new approach for integrating multimodal genetic and imaging data.
• The identified biomarkers and genetic associations provide novel insights into AD pathogenesis.
• This model facilitates the discovery of complex disease association patterns for neurodegenerative disorders.