Review and revamp of compositional data transformation: A new framework combining proportion conversion and contrast...

Area of Science:

• Microbiome research
• Statistical analysis
• Bioinformatics

Background:

• The human microbiome is crucial for health and disease, but its data presents statistical challenges like varying sequencing depth, compositionality, and zero inflation.
• Existing data transformation methods (e.g., scaling, CLR, ALR) can introduce biases and yield inconsistent results.
• Addressing these challenges is vital for accurate microbiome data analysis.

Purpose of the Study:

• To systematically review compositional data transformation techniques for microbiome data.
• To develop a novel framework for creating new data transformations.
• To introduce and evaluate new transformations (CAC, AAC) for zero-inflated microbiome data.

Main Methods:

• Systematic review of existing compositional data transformation methods.
• Development of a new framework combining proportion conversion and contrast transformations.
• Introduction of novel transformations: Centered Arcsine Contrast (CAC) and Additive Arcsine Contrast (AAC).
• Evaluation of transformation performance in scenarios with varying levels of zero-inflation.

Main Results:

• The proposed framework encompasses existing methods like Additive Log Ratio (ALR) and Centered Log Ratio (CLR).
• CAC and AAC transformations demonstrate enhanced performance in high zero-inflation scenarios.
• ALR and CLR transformations are more effective when zero values are less prevalent.
• The study provides a comprehensive comparison of different transformation techniques.

Conclusions:

• The developed framework offers a flexible approach to microbiome data transformation.
• Novel CAC and AAC transformations improve analysis of zero-inflated microbiome datasets.
• The findings guide researchers in selecting appropriate transformation methods for microbiome data analysis, improving analytical outcomes.