Related Experiment Videos
An omnibus test for differential distribution analysis of microbiome sequencing data
Jun Chen1,2, Emily King1,3, Rebecca Deek4
1Division of Biomedical Statistics and Informatics.
Bioinformatics (Oxford, England)
|October 18, 2017
Summary
A new statistical test for microbiome analysis accurately identifies microbial differences by considering abundance, prevalence, and dispersion. This robust method improves upon existing techniques for differential abundance analysis in human microbiome studies.
Area of Science:
- Microbiome research
- Statistical bioinformatics
- Genomic data analysis
Background:
- Human microbiome studies aim to find microbial differences between conditions.
- Existing statistical methods often overlook data dispersion and are sensitive to outliers.
- Current approaches may fail when dispersion varies across samples.
Purpose of the Study:
- To develop a robust statistical method for differential abundance analysis in microbiome data.
- To address limitations of existing methods, including handling outliers and covariate-dependent dispersion.
- To improve the power and reliability of microbiome differential analysis.
Main Methods:
- Introduced a novel test for differential distribution analysis.
- Employed a zero-inflated negative binomial regression model.
- Utilized winsorized count data to manage zero-inflation and outliers.
Main Results:
- The novel test jointly analyzes abundance, prevalence, and dispersion.
- Demonstrated robustness across diverse biological conditions using simulated and real data.
- Showed superior power compared to previous differential abundance methods.
Conclusions:
- The developed method offers a robust and powerful approach for microbiome differential analysis.
- It effectively handles outliers and covariate-dependent dispersion.
- An R package is available for implementation.