Mining Outcome-relevant Brain Imaging Genetic Associations via Three-way Sparse Canonical Correlation Analysis in Alzheimer’s Disease

• Published in: Scientific reports 2017-03, Vol.7 (1), p.44272, Article 44272
• Main Authors: Hao, Xiaoke, Li, Chanxiu, Du, Lei, Yao, Xiaohui, Yan, Jingwen, Risacher, Shannon L., Saykin, Andrew J., Shen, Li, Zhang, Daoqiang
• Format: Article
• Language: English
• Subjects: 631/114/1305; 631/208/721; 631/378/1689; Algorithms; Alzheimer Disease - diagnostic imaging; Alzheimer Disease - genetics; Alzheimer Disease - pathology; Alzheimer's disease; Apolipoprotein E; Apolipoproteins E - genetics; Brain - diagnostic imaging; Brain - metabolism; Brain - pathology; Cognitive ability; Cognitive Dysfunction - diagnostic imaging; Cognitive Dysfunction - genetics; Cognitive Dysfunction - pathology; Cohort Studies; Correlation analysis; Data Mining; Datasets as Topic; Gene Expression; Genetic Association Studies; Genetic markers; Genetic Predisposition to Disease; Genetic variance; Genetics; Humanities and Social Sciences; Humans; Image processing; Magnetic Resonance Imaging; Medical imaging; multidisciplinary; Multivariate Analysis; Neurodegenerative diseases; Neuroimaging; Phenotype; Polymorphism, Single Nucleotide; Science; Severity of Illness Index; Single-nucleotide polymorphism
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/srep44272

Neuroimaging genetics is an emerging field that aims to identify the associations between genetic variants (e.g., single nucleotide polymorphisms (SNPs)) and quantitative traits (QTs) such as brain imaging phenotypes. In recent studies, in order to detect complex multi-SNP-multi-QT associations, bi-multivariate techniques such as various structured sparse canonical correlation analysis (SCCA) algorithms have been proposed and used in imaging genetics studies. However, associations between genetic markers and imaging QTs identified by existing bi-multivariate methods may not be all disease specific. To bridge this gap, we propose an analytical framework, based on three-way sparse canonical correlation analysis (T-SCCA), to explore the intrinsic associations among genetic markers, imaging QTs, and clinical scores of interest. We perform an empirical study using the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort to discover the relationships among SNPs from AD risk gene APOE , imaging QTs extracted from structural magnetic resonance imaging scans, and cognitive and diagnostic outcomes. The proposed T-SCCA model not only outperforms the traditional SCCA method in terms of identifying strong associations, but also discovers robust outcome-relevant imaging genetic patterns, demonstrating its promise for improving disease-related mechanistic understanding.