Bayesian Generalized Low Rank Regression Models for Neuroimaging Phenotypes and Genetic Markers

Area of Science:

• Genetics
• Neuroimaging
• Statistical Modeling

Background:

• High-dimensional data analysis is crucial in genetics and neuroimaging.
• Existing methods may not adequately capture complex relationships between genetic variants and brain phenotypes.
• Alzheimer's Disease Neuroimaging Initiative (ADNI) provides rich data for such analyses.

Purpose of the Study:

• To develop a Bayesian generalized low rank regression model (GLRR) for analyzing high-dimensional responses and covariates.
• To identify associations between genetic variants (SNPs) and brain imaging phenotypes (ROI volumes).
• To provide a robust statistical framework for neuroimaging genetics research.

Main Methods:

• Bayesian generalized low rank regression (GLRR) model.
• Integration of low rank approximation for coefficient matrices.
• Dynamic factor model for high-dimensional covariance matrices.
• Local hypothesis testing for covariate significance.
• Markov chain Monte Carlo (MCMC) for posterior computation.

Main Results:

• The proposed GLRR model effectively analyzes high-dimensional genetic and neuroimaging data.
• Simulation studies demonstrate the performance of GLRR compared to existing methods.
• Application to ADNI data reveals associations between specific SNPs and brain region volumes.

Conclusions:

• GLRR offers a powerful tool for exploring complex associations in neuroimaging genetics.
• The model facilitates the identification of genetic influences on brain structure relevant to diseases like Alzheimer's.
• This approach enhances our understanding of the genetic underpinnings of brain phenotypes.