Methods for mediation analysis with high-dimensional DNA methylation data: Possible choices and comparisons.

Area of Science:

• Epigenetics
• Statistical Genetics
• Environmental Health

Background:

• DNA methylation is frequently studied as a mediator linking environmental exposures to health outcomes.
• Existing statistical methods for high-dimensional mediation analysis are not widely adopted in epigenetics research.
• There is a need for robust methods to analyze complex epigenetic mediation pathways.

Purpose of the Study:

• To compare the performance of seven statistical methods for high-dimensional mediation analysis with continuous outcomes.
• To identify the best-performing methods for detecting active mediators and estimating global mediation effects.
• To provide practical guidelines and an R package for implementing these methods in epigenetic research.

Main Methods:

• Simulation studies were conducted to evaluate method performance under various scenarios.
• Real-world DNA methylation data from a large, multi-ethnic US cohort were analyzed.
• Seven distinct high-dimensional mediation analysis methods were compared, including BSLMM, HDMA, HILMA, and PCMA.

Main Results:

• Bayesian sparse linear mixed model (BSLMM) and high-dimensional mediation analysis (HDMA) showed superior performance in detecting active mediators in simulations.
• High-dimensional linear mediation analysis (HILMA) and principal component mediation analysis (PCMA) were preferred for estimating the global mediation effect.
• The study identified specific methods that excel in different aspects of mediation analysis.

Conclusions:

• Guidelines are provided for epigenetic researchers to select appropriate high-dimensional mediation analysis methods based on study objectives.
• The findings facilitate the adoption of advanced statistical techniques in the analysis of DNA methylation data.
• Future research directions for methodological development in mediation analysis are suggested.