A region-based method for causal mediation analysis of DNA methylation data

Area of Science:

• Epigenetics
• Environmental Health
• Biostatistics

Background:

• Environmental factors influence health outcomes through epigenetic modifications like DNA methylation.
• Single CpG site analysis is common, but differentially methylated regions (DMRs) may have greater functional impact and reduce multiple testing.
• Causal mediation analysis is crucial for understanding how environmental exposures affect outcomes via DNA methylation.

Purpose of the Study:

• To propose a novel causal mediation approach (MR-Mediation) for analyzing the effects of predictors on outcomes mediated by methylated regions (MRs).
• To test total (TE), direct (DE), and indirect effects (IE) using a counterfactual framework.
• To evaluate the performance of MR-Mediation compared to traditional methods.

Main Methods:

• Developed MR-Mediation, a causal mediation approach using a counterfactual framework.
• Employed functional linear transformation to handle high-dimensional CpG sites within MRs and incorporate location information.
• Conducted simulation studies to assess Type I error rates and statistical power.

Main Results:

• MR-Mediation maintained desired Type I error rates for TE, DE, and IE tests.
• The method demonstrated superior power compared to analyzing mean methylation levels, particularly for indirect effects (IE).
• Applied MR-Mediation to analyze the impact of gun violence exposure on immunoglobulin E and atopic asthma, mediated by DNA methylation.

Conclusions:

• MR-Mediation provides a robust framework for assessing causal mediation through methylated regions.
• The approach enhances the understanding of environmental epigenetics and its role in disease.
• This method offers improved power for detecting mediated effects, especially indirect ones, in complex biological systems.