Analysis of composition of microbiomes: a novel method for studying microbial composition

Area of Science:

• Microbiome research
• Statistical methodology
• Bioinformatics

Background:

• Understanding microbiota composition is crucial but requires robust statistical methods.
• Existing approaches often overlook microbiome data's compositional nature or use unsuitable probability models.
• This can lead to inaccurate conclusions in comparative microbiome studies.

Purpose of the Study:

• To develop a statistical framework that respects compositional data constraints.
• The goal is to reduce false discoveries in detecting differentially abundant taxa.
• Maintain high statistical power in microbiome analyses.

Main Methods:

• Introduced the Analysis of Composition of Microbiomes (ANCOM) statistical framework.
• ANCOM accounts for data structure without making distributional assumptions.
• It integrates with linear models for covariate adjustment and longitudinal data, scaling to thousands of taxa.

Main Results:

• ANCOM demonstrated superior performance over t-tests and Zero Inflated Gaussian (ZIG) methods.
• ANCOM effectively controlled the false discovery rate (FDR) and enhanced statistical power.
• Standard methods like t-test and ZIG showed inflated FDRs, reaching up to 68% and 60% respectively.

Conclusions:

• Accounting for the compositional nature of microbiome data is essential for accurate inference.
• Log-ratio analysis within ANCOM significantly improves the interpretation of microbiota survey data.
• ANCOM provides a powerful and reliable tool for microbiome comparative studies.