Kernel Method for Detecting Higher Order Interactions in multi-view Data: An Application to Imaging, Genetics, and Epigenetics

摘要: In this study, we tested the interaction effect of multimodal datasets using a novel method called kernel for detecting higher order interactions among biologically relevant mulit-view data. Using semiparametric on reproducing Hilbert space (RKHS), used standard mixed-effects linear model and derived score-based variance component statistic that tests between multi-view The proposed offers an intangible framework identification effects (e.g., three way interaction) genetics, brain imaging, epigenetic Extensive numerical simulation studies were first conducted to evaluate performance method. Finally, was evaluated data from Mind Clinical Imaging Consortium (MCIC) including single nucleotide polymorphism (SNP) data, functional magnetic resonance imaging (fMRI) scans, deoxyribonucleic acid (DNA) methylation respectfully, in schizophrenia patients healthy controls. We treated each gene-derived SNPs, region interest (ROI) DNA as testing unit, which are combined into triplets evaluation. addition, cardiovascular disease risk factors such age, gender, body mass index assessed covariates hippocampal volume compared triplets. Our identified $13$-triplets ($p$-values $\leq 0.001$) included $6$ $10$ ROIs, methylations correlated with changes volume, suggesting these may be important explaining schizophrenia-related neurodegeneration. With strong evidence 0.000001$), triplet ({\bf MAGI2, CRBLCrus1.L, FBXO28}) has potential distinguish control variations.