An Efficient Testing Procedure for High-Dimensional Mediators with FDR Control

The field of mediation analysis commonly explores the pathways that connect environmental exposures with health outcomes. With the development of data collection techniques, greater efforts have been dedicated to addressing high-dimensional mediators. In this paper, we present an efficient approach to identify significant mediators while controlling the false discovery rate (FDR). We propose a three-step procedure that incorporates independent screening, variable selection together with refitted partial regression, and divide-aggregate composite-null test (DACT). The simulation includes a comparative analysis of our proposed method in comparison to eight competing approaches, demonstrating that our procedure has significant advantages over other methods. The proposed procedure is applied to investigate the mediation mechanisms of DNA methylation in the relationship between smoking and lung function. Three specific methylation sites (cg26331243, cg19862839, and cg12616487) are identified as potential epigenetic markers involved in mediating this relationship. Our proposed method is available with the R package HIMA at https://cran.r-project.org/web/packages/HIMA/.