Group-wise normalization in differential abundance analysis of microbiome samples.

Area of Science:

• Microbiology
• Bioinformatics
• Statistical analysis

Background:

• Microbial sample counts are compositional, leading to biased differential abundance analysis.
• Existing normalization methods struggle to maintain false discovery rates with high variance or compositional bias.

Purpose of the Study:

• Propose a novel group-level normalization framework to reduce bias in differential abundance analysis.
• Introduce two new normalization methods: group-wise relative log expression (G-RLE) and fold-truncated sum scaling (FTSS).

Main Methods:

• Re-conceptualized normalization as a group-level task.
• Developed and applied G-RLE and FTSS normalization methods.
• Evaluated methods using model-based and synthetic data simulations.

Main Results:

• G-RLE and FTSS demonstrated higher statistical power in identifying differentially abundant taxa compared to existing methods.
• The proposed methods maintained the false discovery rate in challenging scenarios.
• FTSS normalization combined with MetagenomeSeq yielded the best performance.

Conclusions:

• The group-wise normalization framework effectively reduces bias in microbial differential abundance analysis.
• FTSS normalization is a powerful tool for analyzing compositional microbial data.
• The developed methods offer improved accuracy and reliability in microbiome research.