Project Summary/Abstract
Alzheimer's disease (AD) is one of the leading causes of death for the elderly with no current cure. Genomics
studies, such as mapping expression quantitative trait loci (eQTL) and differential gene expressions, play a
critical role in understanding the biological mechanisms of AD and developing potential therapeutic treatments.
In genomics studies, there has been growing awareness that the covariates (e.g., quantitative gene expression)
may have changing effects on the distribution of responses (e.g., disease phenotypes) re¿ecting a heterogeneous
covariates-response association. Those heterogeneous associations shed insight on scienti¿c discoveries and
entail signi¿cant implications but are often neglected by most existing analysis procedures con¿ned to a narrow
aspect of the response distribution (e.g., standard linear regression focusing on the mean or quantile regression
at a single quantile level). Thus, the development of valid and ef¿cient hypothesis tests to detect heterogeneous
associations is of great value to genomics studies of complex diseases such as AD.
This proposal aims to develop several quantile regression-based global signi¿cance tests, which utilize all
information across a well-chosen region of quantile levels and provide researchers with evaluations of the overall
impacts of covariates on the response. Inspired by our preliminary data analysis on the two studies of aging and
dementia, namely Religious Orders Study (ROS) and Memory and Aging Project (MAP), we will ¿rst propose
a global signi¿cance test to thoroughly evaluate covariates' impact across all quantile levels of the response
variable (Aim 1). Then motivated by high-dimensional genomics data of AD in ROS/MAP, we will further develop
two global signi¿cance tests for high-dimensional responses and covariates data, respectively (Aim 2). Moreover,
we will apply the proposed tests in Aims 1-2 to the genomics data generated by ROS/MAP to identify eQTL and
differentially expressed genes that can be used to prioritize risk genes of AD for identifying developing potential
treatments (Aim 3). We will also provide a user-friendly R package to implement the proposed tests.
The innovation of our proposal is three-fold. (i) By evaluating the impacts of covariates on responses across
the entire quantile domain, the proposed global signi¿cance tests have a superior power to identify heterogeneous
covariates-response associations compared to alternative methods. (ii) As the proposed tests neither impose any
stringent model assumption nor require additional splines smoothing or re-sampling or shrinkage estimation, they
can be broadly implemented in large-scale genomics data. (iii) Our proposed test in Aim 2 will serve as a useful
tool for detecting heterogeneous associations between covariates and multiple responses.
The successful completion of this project will facilitate detecting heterogeneous associations and the
subsequent scienti¿c discoveries in AD genomics studies for developing treatments. Moreover, our tests can
be applied to a broad scope of biomedical ¿elds, resulting in a fruitful avenue for promoting public health.