Robust multivariate regression controlling false discoveries for microbiome data

Area of Science:

• Microbiome research
• Statistical modeling
• Bioinformatics

Background:

• Microbiome signatures are crucial for understanding diseases like obesity and liver disease.
• Analyzing microbiome data presents challenges due to compositionality, high dimensionality, sparsity, and outliers.

Purpose of the Study:

• To develop a robust multivariate compositional regression model for identifying microbiome-health indicator associations.
• To address the limitations of existing methods in analyzing complex microbiome data.

Main Methods:

• Developed a robust multivariate compositional regression model.
• Incorporated outlier robustness and a derandomization step.
• Ensured control of the false discovery rate (FDR) for reliable results.

Main Results:

• The proposed method outperforms the Multi-Response Knockoff Filter (MRKF) in simulation studies regarding FDR control, power, and robustness.
• Successfully identified microbial species associated with specific clinical parameters in real-world data applications.
• Enhanced stability and reproducibility of microbiome data analysis.

Conclusions:

• The developed robust regression model offers a superior approach for analyzing microbiome data and discovering disease-associated microbial signatures.
• Provides valuable biological insights by reliably linking microbial species to clinical health indicators.
• The method is available as R code with comprehensive documentation.