Estimating and testing the microbial causal mediation effect with high-dimensional and compositional microbiome data.

Area of Science:

• Microbiome research
• Causal inference
• Statistical genetics

Background:

• Microbiome association studies highlight links between microbial composition and health status.
• Investigating the causal role of the microbiome is crucial for understanding biological mechanisms.
• Existing causal mediation methods are inadequate for high-dimensional, compositional microbiome data.

Purpose of the Study:

• To propose a novel causal mediation model tailored for microbiome data.
• To enable the identification of specific microbial agents influencing health outcomes.
• To quantify causal microbiome effects in a three-factor study design.

Main Methods:

• Developed the Sparse Microbial Causal Mediation Model (SparseMCMM).
• Utilized linear log-contrast and Dirichlet regression for effect estimation.
• Employed regularization techniques for variable selection and identification of causal microbes.
• Proposed hypothesis tests for overall mediation effects with permutation-based significance estimation.

Main Results:

• SparseMCMM demonstrates excellent performance in estimation and hypothesis testing via simulations.
• The model successfully identified causal pathways in a murine antibiotic treatment study.
• Demonstrated utility in linking antibiotic treatment, microbiome composition, and mouse weight.

Conclusions:

• SparseMCMM provides a rigorous framework for causal microbiome analysis.
• The method effectively identifies key microbial mediators in complex biological systems.
• Offers a valuable tool for advancing microbiome research and understanding host-microbe interactions.