Low-Rank Matrix Regression Models for Large-Scale Imaging Genetic Data

• Activity Number: 249
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Dehan Kong* and Baiguo An and Hongtu Zhu
• Companies: The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill
• Keywords: ADNI ; Low-Rank ; Imaging-Genetics

Abstract

Jointly analyzing high-dimensional imaging measures and genetic variations poses significant computational challenges, including limited computer memory, finite CPU speed, and limited CPU nodes. The aim of this paper is to develop low-rank matrix regression models to efficiently carry out whole-genome analyses of matrix response data. Our matrix regression model explicitly treats the matrix response data as a matrix response instead of vectorizing it as a multivariate dimensional vector. Our procedure can eliminate a large number of 'noisy' genetic markers, while preserving a small set of important markers. We can also estimate matrix coefficient images, while explicitly exploiting the spatial structure of matrix responses. We examine the finite sample performance of our methods by using simulations and the imaging genetic data from the Alzheimer's Disease Neuroimaging Initiative. Asymptotic theories have also been developed for our procedures.