ELX: An Empirical Two-Stage Combinatorial Approach to Identify Pleiotropic Genetic Effects by using Genome-Wide Association Summary Statistics

Abstract Genome-wide association studies (GWAS) have successfully identified genetic determinants underlying complex human diseases. Pleiotropic effects occur when a single genetic variant influences multiple phenotypes or illnesses, widely observed in recent GWAS findings. However, no systematic effort has been made to statistically detect pleiotropic effects across many traits/phenotypes due to the lack of easy-to-use statistical tools utilizing summary statistics from GWAS, mainly since most GWAS were conducted via meta-analyzing combining summary statistics from multiple studies. This study proposes an Empirical Linear two-stage Combinatorial (ELX) method to identify potential pleiotropic effects by utilizing aggregated results from GWAS summary statistics. In the first stage, we developed direct linear combining (dLC), and empirical combining (eLC) approaches combining correlated univariate test statistics to screen potential pleiotropic variants on a genome-wide scale. In the second stage, we developed a conditional pleiotropy testing (cPLT) approach to examine the pleiotropic effects for candidate variants identified in Stage 1 using individual-level data. The results demonstrated that cPLT reduced type 1 error in identifying pleiotropic genetic variants compared to the typical conditional strategy. We validated ELX by performing a bivariate GWA study on two correlated quantitative traits, high-density lipoprotein and triglycerides, in the Framingham Heart Study. The proposed two-stage approach allows us to leverage aggregated summary statistics from univariate GWAS. It improves the power to identify potential pleiotropy while maintaining low false-positive rates. ELX is available on Github at https://yihsianghsulab.github.io/ELX.