Power-enhanced two-sample mean tests for high-dimensional microbiome compositional data.

Area of Science:

• Microbiome research
• Statistical analysis
• High-dimensional data

Background:

• Comparing microbial communities is crucial for understanding their function.
• Current statistical methods may lack power for certain microbiome data patterns.
• High-dimensional compositional data presents unique analytical challenges.

Purpose of the Study:

• To develop a novel 2-sample mean test for high-dimensional compositional microbiome data.
• To enhance statistical testing power and robustness across diverse signal patterns.
• To improve the detection of differences in microbial community structures.

Main Methods:

• Developed a power-enhanced mean test by combining P-values from maximum-type and quadratic-type tests.
• Integrated strengths of existing popular statistical tests.
• Provided theoretical guarantees for Type-I error control and power enhancement.

Main Results:

• The proposed test demonstrates accurate Type-I error rate control.
• Achieved significantly enhanced testing power across a broad range of alternative hypotheses.
• Showcased robust performance in both simulated and real-world microbiome datasets.

Conclusions:

• The novel power-enhanced mean test offers substantial improvements over existing methods for microbiome data.
• The methodology contributes to advancements in high-dimensional hypothesis testing and power enhancement.
• This approach provides a more reliable tool for microbiome compositional data analysis.