A Transformation‐Based Direction Combination Association Test for <scp>GWAS</scp> Summary Statistics

ABSTRACT Genome‐wide association study (GWAS) has identified many genetic variants associated with complex diseases. Traditionally, GWAS tests the association between a single variant and a single trait. Gene‐based methods utilize the fact that multiple SNPs function as a gene to affect traits, extending GWAS to test the association between a trait and multiple variants. Nowadays, only summary statistics instead of individual‐level data from GWAS are publicly available due to privacy limits. Existing summary statistics for gene‐based methods may lead to power loss under the opposite signs of effect coefficients. In this paper, we develop a novel gene‐based transformed test addressing this defect. Instead of directly building test statistics using Wald test statistics, we transform Wald test statistics back into true effect coefficients based on their relationship. Then, we propose an ensemble test to produce robustness under various alternative hypotheses. Extensive simulations show that our proposed method demonstrates high power compared to existing methods. Analysis of real data on polyunsaturated fatty acids shows that our method can identify additional genetic markers previously undetected by other methods.