High Dimensional Mediation Analysis: a new method applied to maternal smoking, placental DNA methylation and birth outcomes

Abstract

Most high dimensional mediation epigenetic studies evaluate each mediator indirect effects independently. These approaches are underpowered to detect multiple mediators and the overall indirect effect remains poorly quantified. We developed HDMAX2, an efficient algorithm for high dimensional mediation analysis using max-squared tests, considering CpGs and aggregated mediator regions. HDMAX2 outperformed existing approaches in simulated and real data, detected true mediators with increased power, and assessed the overall indirect effect of several mediators from a high dimension matrix. We showed a polygenic architecture for placenta CpGs and regions mediating the effects of maternal smoking during pregnancy on babys birth weight.