A comprehensive evaluation of microbial differential abundance analysis methods: current status and potential...

Area of Science:

• Microbiome research
• Bioinformatics
• Statistical analysis

Background:

• Differential abundance analysis (DAA) is crucial for microbiome studies to identify microbial biomarkers.
• Existing DAA tools face challenges with zero-inflation and compositional effects, leading to inconsistent results.
• A comprehensive evaluation of DAA tools is needed to guide best practices.

Purpose of the Study:

• To comprehensively evaluate existing differential abundance analysis (DAA) tools for microbiome data.
• To identify the strengths and weaknesses of current DAA methods across various biologically relevant scenarios.
• To develop an improved DAA tool that addresses the limitations of existing methods.

Main Methods:

• Performed a comprehensive evaluation of existing DAA tools using real data-based simulations.
• Assessed performance based on false-positive control and statistical power.
• Developed ZicoSeq, an optimized procedure integrating strengths of existing DAA methods.

Main Results:

• DAA methods addressing compositional effects showed improved false-positive control but were not optimal.
• The LDM method offered high power but had unsatisfactory false-positive control in certain settings.
• ZicoSeq demonstrated robust false-positive control across settings with high statistical power, validated on real datasets.

Conclusions:

• No single existing DAA tool is universally applicable to all microbiome datasets.
• The performance of DAA tools is highly dependent on unknown a priori settings.
• ZicoSeq is a robust and flexible DAA tool for microbiome biomarker discovery, overcoming limitations of existing methods.