Genome-wide multimediator analyses using the generalized Berk-Jones statistics with the composite test

Area of Science:

• Genomics
• Statistical genetics
• Bioinformatics

Background:

• Mediation analysis assesses causal pathways from exposure to outcome via mediators.
• High-throughput technologies generate large-scale genomic and proteomic data, necessitating robust statistical methods.
• Conventional mediation analysis methods lack statistical power for sparse effects under composite null hypotheses, leading to missed discoveries.

Purpose of the Study:

• To extend composite null hypothesis testing for mediation analysis to multimediator scenarios.
• To develop an R package (MACtest) for efficient and powerful multimediator analysis in genomic studies.
• To identify novel genes and networks regulated by smoking-induced epigenetic aberrations.

Main Methods:

• Utilized generalized Berk-Jones statistics combined with a composite test framework.
• Developed three approaches: multivariate (dense/diverse effects), decorrelation (sparse/consistent effects), and hybrid.
• Implemented the methods in the R package MACtest.

Main Results:

• The MACtest package provides adjusted P-values for multimediator analyses, enhancing power for sparse effects.
• Analysis of lung adenocarcinoma datasets identified potential genes and networks regulated by smoking-induced epigenetic aberrations.
• Demonstrated the utility of MACtest in real-world genomic and proteomic studies.

Conclusions:

• MACtest offers a powerful and flexible tool for multimediator analysis in the era of high-throughput genomics.
• The developed methods improve the detection of true causal mechanisms, especially under sparse effect scenarios.
• The package facilitates the investigation of complex biological questions, such as epigenetic regulation by environmental factors.