Identify connectome between genotypes and brain network phenotypes via deep self-reconstruction sparse canonical correlation analysis

• DOI: 10.1093/bioinformatics/btac074
• Publication Date: 2022-02-02T20:23:48Z
• Authors: Meiling Wang; Wei Shao; Xiaoke Hao; Shuo Huang; Daoqiang Zhang
• Keywords: Imaging genetics; Human Connectome Project; Connectomics

Abstract Motivation As a rising research topic, brain imaging genetics aims to investigate the potential genetic architecture of both structure and function. It should be noted that in brain, not all variations are deservedly caused by effect, it is generally unknown which phenotypes promising for analysis. Results In this work, variants (i.e. single nucleotide polymorphism, SNP) can correlated with networks quantitative trait, QT), so connectome (including regions connectivity features) functional from magnetic resonance data identified. Specifically, connection matrix firstly constructed, whose upper triangle elements selected features. Then, PageRank algorithm exploited estimating importance different as region Finally, deep self-reconstruction sparse canonical correlation analysis (DS-SCCA) method developed identification associations phenotypic markers. This approach regularized, extension, scalable multi-SNP-multi-QT method, well-suited applying association Alzheimer’s Disease Neuroimaging Initiative datasets. further optimized adopting parametric approach, augmented Lagrange stochastic gradient descent. Extensive experiments provided validate DS-SCCA realizes strong discovers biomarkers guide disease interpretation. Availability implementation The Matlab code available at https://github.com/meimeiling/DS-SCCA/tree/main. Supplementary information Bioinformatics online.