Generalized Hotelling's test for paired compositional data with application to human microbiome studies.

Area of Science:

• Microbiology
• Statistical Bioinformatics
• Human Microbiome Research

Background:

• The human microbiome is dynamic and influenced by various factors like disease, diet, and medication.
• Paired study designs are crucial for analyzing microbiome changes within individuals, controlling for personal variations.
• Analyzing microbiome data presents challenges due to its compositional nature and high dimensionality relative to sample size.

Purpose of the Study:

• To develop a robust statistical method for analyzing paired microbiome data.
• To address the compositional nature and high dimensionality of microbiome datasets.
• To evaluate the equivalence of average microbiome compositions in paired samples.

Main Methods:

• Log-ratio transformation of compositional microbiome data.
• Development of a generalized Hotelling's test (GHT) using shrinkage-based covariance estimation.
• Implementation of a permutation procedure for statistical significance assessment.

Main Results:

• The proposed GHT method demonstrates well-controlled Type I error rates in simulations.
• The GHT method exhibits superior statistical power compared to existing ad hoc approaches.
• The method was successfully applied to analyze vaginal microbiome shifts in response to menopausal hot flash treatments.

Conclusions:

• The generalized Hotelling's test (GHT) provides a powerful and reliable approach for analyzing paired microbiome data.
• The method effectively handles the unique challenges of compositional and high-dimensional microbiome datasets.
• The developed R package "GHT" is available for researchers studying microbiome dynamics.