Mining high-level brain imaging genetic associations

• Main Author: Yao, Xiaohui
• Other Authors: Wu, Huanmei
• Language: en_US
• Published: 2018
• Subjects: Alzheimer's disease; Enrichment analysis; Imaging genetics; Machine learning; Tissue-specific; Interaction network
• Online Access: http://hdl.handle.net/1805/15831

Indiana University-Purdue University Indianapolis (IUPUI) === Imaging genetics is an emerging research field in neurodegenerative diseases. It studies the influence of genetic variants on brain structure and function. Genome-wide association studies (GWAS) of brain imaging has identified a few independent risk loci for individual imaging quantitative trait (iQT), which however display only modest effect size and explain limited heritability. This thesis focuses on mining high-level imaging genetic associations and their applications on neurodegenerative diseases. This thesis first presents a novel network-based GWAS framework for identifying functional modules, by employing a two-step strategy in a top-down manner. It first integrates tissue-specific network with GWAS of corresponding phenotype in regression models in addition to classification, to re-prioritize genome-wide associations. Then it detects densely connected and disease-relevant modules based on interactions among top reprioritizations. The discovered modules hold both phenotypical specificity and densely interaction. We applied it to an amygdala imaging genetics analysis in the study of Alzheimer's disease (AD). The proposed framework effectively detects densely interacted modules; and the reprioritizations achieve highest concordance with AD genes. We then present an extension of the above framework, named GWAS top-neighbor-based (tnGWAS); and compare it with previous approaches. This tnGWAS extracts densely connected modules from top GWAS findings, based on the hypothesis that relevant modules consist of top GWAS findings and their close neighbors. It is applied to a hippocampus imaging genetics analysis in AD research, and yields the densest interactions among top candidate genes. Experimental results demonstrate that precise context does help explore collective effects of genes with functional interactions specific to the studied phenotype. In the second part, a novel imaging genetic enrichment analysis (IGEA) paradigm is proposed for discovering complex associations among genetic modules and brain circuits. In addition to genetic modules, brain regions of interest also grouped to play role. We expand the scope of one-dimensional enrichment analysis into imaging genetics. This framework jointly considers meaningful gene sets (GS) and brain circuits (BC), and examines whether given GS-BC module is enriched in gene-iQT findings. We conduct the proof-of-concept study and demonstrate its performance by applying to a brain-wide imaging genetics study of AD.