A Bottom-up Approach to Testing Hypotheses That Have a Branching Tree Dependence Structure, with Error Rate Control

Area of Science:

• Microbiology
• Statistical Genetics
• Bioinformatics

Background:

• Modern statistical analyses frequently involve testing numerous hypotheses, often with inherent dependencies.
• Microbiome research often organizes microbial data into hierarchical taxonomic structures (e.g., species, genus).
• Testing associations within these hierarchical structures requires accounting for dependencies between tests.

Purpose of the Study:

• To develop a novel bottom-up statistical testing algorithm for hierarchical microbiome data.
• To control a new error rate termed the false selection rate.
• To improve the identification of "driver taxa"—higher taxonomic groups with dense association signals.

Main Methods:

• Development of a bottom-up hypothesis testing algorithm starting from operational taxonomic units (OTUs) or amplicon sequence variants (ASVs).
• The algorithm proceeds upwards through taxonomic levels (species, genus, family, etc.).
• Control of a novel "false selection rate" to manage statistical errors.

Main Results:

• The developed algorithm effectively controls the false selection rate.
• Simulations demonstrate superior performance in identifying driver taxa compared to existing methods.
• The approach was illustrated using microbiome data from ulcerative colitis patients and healthy controls.

Conclusions:

• The novel bottom-up testing algorithm provides a robust method for analyzing hierarchical microbiome data.
• This approach enhances the ability to detect significant associations at various taxonomic levels.
• The method is particularly useful for identifying key microbial groups driving trait associations.