Aggregating Knockoffs for False Discovery Rate Control with an Application to Gut Microbiome Data

Area of Science:

• Microbiome research
• Statistical genetics
• Bioinformatics

Background:

• The gut microbiome significantly impacts human health and well-being.
• High-dimensional microbiome data present analytical challenges, including the risk of false discoveries with traditional methods like the lasso estimator.
• Knockoffs offer a recent statistical approach to control false discoveries.

Purpose of the Study:

• To develop and validate an aggregated knockoff method for enhanced power in high-dimensional microbiome data analysis.
• To rigorously control the number of false discoveries in microbiome association studies.
• To identify novel microbial associations with health conditions, specifically obesity.

Main Methods:

• Development of an aggregated knockoff procedure for high-dimensional data.
• Theoretical validation of the proposed method.
• Simulation studies to assess performance and power.
• Application to real-world microbiome data from the American Gut Project.

Main Results:

• The aggregated knockoff method increases statistical power while maintaining strict control over false discoveries.
• Simulations confirm the theoretical guarantees and practical utility of the approach.
• Analysis of American Gut Project data revealed significant associations between previously overlooked gut phyla and obesity.

Conclusions:

• Aggregated knockoffs provide a powerful and reliable tool for analyzing complex microbiome data.
• This method facilitates the discovery of novel microbial biomarkers for health and disease.
• The findings highlight potential new targets for understanding and managing obesity through gut microbiome modulation.