Abstract:
|
In genome-wise association studies (GWAS), inferring the significance of a large number of genomic markers on a clinical phenotype is hindered by the multiple testing problem, in which controlling Type I error significantly deteriorates the power of tests. Test power can be improved by augmenting signal-to-noise ratio through auxiliary covariates strongly associated with the phenotype. For example, brain imaging data can serve to improve such inference for phenotypes related to mental illness or cognitive ability. In this paper, we develop a systematic method for reducing dimension size in auxiliary data that reduces signal-to-noise ratio in GWAS yielding more powerful tests. After applying principal component analysis on imaging data from a given number of subjects, our method estimates factors for each subject based on thresholded factor loadings. These factors are then employed as covariates while we assess the association genomic markers with the phenotype. We apply this method to the data from the Pediatric, Imagining, and Neurocognitive (PING) study to infer associations between genomic markers, anatomical regions in the brain, and cognitive traits.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.