ADAPT: Analysis of Microbiome Differential Abundance by Pooling Tobit Models

Area of Science:

• Microbiome research
• Bioinformatics
• Statistical modeling

Background:

• Differential abundance analysis (DAA) of microbiome data is challenging due to excessive zeros and data compositionality.
• Existing DAA methods often make unfounded assumptions or employ complex models to address these issues.

Purpose of the Study:

• To introduce a novel method, ADAPT (Analysis of Microbiome Differential Abundance by Pooling Tobit Models), designed to overcome challenges in microbiome DAA.
• To provide a robust statistical framework for identifying differentially abundant microbial taxa.

Main Methods:

• ADAPT treats zero counts as left-censored observations, avoiding complex modeling.
• It incorporates a theoretically sound approach for selecting non-differentially abundant taxa as a reference.
• The method is implemented as an R package available on Bioconductor and GitHub.

Main Results:

• Simulations demonstrate ADAPT offers superior false discovery rate control and higher statistical power compared to existing methods.
• Application to 16S rRNA and shotgun metagenomics data from the COHRA2 study revealed novel insights.
• Analysis of infant oral microbiome data identified associations with early childhood dental caries.

Conclusions:

• ADAPT provides a robust and accurate solution for microbiome differential abundance analysis.
• The method enhances the understanding of microbial community structure and its relation to health outcomes.
• ADAPT facilitates more reliable discoveries in microbiome research, particularly in complex datasets.