CBEA: Competitive balances for taxonomic enrichment analysis

Area of Science:

• Microbiome research
• Bioinformatics
• Statistical genomics

Background:

• Human microbiome research relies on high-throughput sequencing, generating complex, high-dimensional, sparse, and compositional taxonomic count data.
• Existing statistical methods struggle with microbiome data complexity, and current set-based aggregation techniques like abundance summation have limitations.
• There's a need for sophisticated set-based analysis methods tailored for microbiome taxonomic data to enhance interpretability and analytical power.

Purpose of the Study:

• To develop a novel single-sample taxon enrichment method for microbiome data.
• To address limitations of current aggregation techniques by proposing a new log-ratio formulation.
• To provide a robust method for analyzing high-dimensional, sparse, and compositional microbiome data.

Main Methods:

• Developed Competitive Balances for Taxonomic Enrichment Analysis (CBEA), a single-sample method.
• Utilized a novel log-ratio formulation based on the competitive null hypothesis.
• Generated sample-specific enrichment scores as scaled log-ratios of within-set and complement subcompositions.
• Implemented sample-level significance testing via empirical null distribution estimation.

Main Results:

• CBEA effectively controls for type I error, even with high sparsity and inter-taxa correlation.
• The method generates informative, sample-specific enrichment scores.
• Demonstrated validity and performance through real data applications and simulations.
• CBEA scores are suitable for downstream analyses, including prediction tasks.

Conclusions:

• CBEA offers a significant advancement in microbiome data analysis, overcoming limitations of existing methods.
• The method enhances interpretability and statistical power in microbiome research.
• CBEA provides a reliable tool for generating robust enrichment scores and facilitating downstream applications.