Multiple-Scale Adaptive Mixed Model for Imaging Genetics Data

• Activity Number: 77
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Ja-an Lin*+ and Hongtu Zhu and Wei Sun and Joseph Ibrahim
• Companies: and 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: Imaging genetics ; mixed model ; adaptive weights ; voxel-wise method ; multiscale adaptive regression

Abstract

Traditional statistical methods analyzing imaging genetics data, which includes medical imaging measurements (e.g. MRI) and genomic information (e.g. SNP), often suffer from low statistical power. We propose a method named Multiple-scale Adaptive Mixed Model (MAMM). It incorporates the genomic information as a population-shared random effect with a common variance component (VC), and other clinical variables as fixed effect in a weighted likelihood model. Then we adaptively include neighbourhood voxels with a certain weight while estimating both the VC and the regression coefficients. The integrated genomic effect is investigated by testing the zero of the VC via weighted likelihood ratio test with similar adaptive strategy involving nearby voxels. The performance of MAMM is evaluated by simulation studies and the result shows it outperforms voxel-wise based approach by greater statistical power with great type I error control. MAMM is also applied to the ADNI study to search possible SNPs associating to the Alzheimer' disease progression.