In this article, we propose a unified statistical framework, named GPA (Genetic analysis incorporating Pleiotropy and Annotation), to prioritize GWAS results based on pleiotropy and annotation information. GPA also provides statistically rigorous and biologically interpretable inference tools for this purpose. The method can easily be used for other purposes as well, as we will discuss below. This article is organized as follows. First, we investigate the properties of GPA using extensive simulation studies and illustrate the versatility and utility of GPA with the analysis of real data. Specifically, we apply GPA to the five psychiatric disorder GWAS data from the Psychiatric Genomics Consortium with central nervous system gene expression data and show that GPA can accurately identify pleiotropy structure among these diseases. We further apply GPA to the bladder cancer GWAS data with the ENCODE DNase-seq data from 125 cell lines and show that GPA can detect cell lines that are biologically more relevant to bladder cancer. Lastly, we discuss many issues related to GPA. The details of our GPA model and its statistical inference procedures are provided in the Materials and Methods section.
