CRAmed: a conditional randomization test for high-dimensional mediation analysis in sparse microbiome data

Area of Science:

• Microbiome research
• Statistical genetics
• Computational biology

Background:

• Microbiome studies show links to human health and disease.
• Understanding the microbiome's causal role in complex traits is crucial.
• Microbiome data complexity challenges causal effect analysis.

Purpose of the Study:

• Introduce CRAmed, a novel statistical framework for microbiome mediation analysis.
• Improve interpretability of mediation analysis by decomposing effects based on microbial presence-absence and abundance.
• Evaluate CRAmed's performance against existing methods via simulations and real data.

Main Methods:

• Developed a statistical framework named CRAmed.
• Implemented mediation analysis decomposing the natural indirect effect into microbial presence-absence and abundance components.
• Conducted comprehensive simulations and applied the method to two real-world datasets.

Main Results:

• CRAmed demonstrated superior performance in Recall, precision, and F1 score compared to existing methods.
• The framework showed robustness in simulations.
• Real data applications confirmed CRAmed's effectiveness and interpretability in uncovering microbiome's mediating role.

Conclusions:

• CRAmed offers a promising approach for investigating the microbiome's mediating role in health and disease.
• The method enhances understanding of factors influencing host health through microbial interactions.
• The R package CRAmed is publicly available for broader research application.