# Testing for differential abundance in compositional counts data, with application to microbiome studies

**Authors:** Barak Brill, Amnon Amir, Ruth Heller  
**arXiv:** [1904.08937](https://arxiv.org/abs/1904.08937)  
**Date:** 2019-04-18  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/1904.08937v1/

## Key points

- Study focuses on identifying differences in microbiome community across different groups
- Measurement of relative frequencies of taxa using sequencing PCR amplicons
- Statistical inference is challenging due to high number of taxa and strong correlations between them
- Data is compositional and sparse with technical zeros present
- Proposed novel approach for differential abundance testing using a set of reference taxa and data-adaptive method for identifying them
- Existing methods do not provide control over false positive discoveries or valid inference in certain scenarios
- Valuable contribution to the field by addressing limitations of existing methods and providing new approach for analyzing compositional counts data with technical zeros

## Summary

The study focuses on identifying differences in microbiome community across different groups by measuring relative frequencies of taxa using sequencing PCR amplicons. However, statistical inference is challenging due to a high number of taxa compared to sampled units and strong correlations between them. The total number of sequenced reads per sample is limited by the sequencing procedure, making the data compositional and sparse with technical zeros present. To address these challenges, researchers propose a novel approach for differential abundance testing using a set of reference taxa and a data-adaptive method for identifying them. Existing methods do not provide control over false positive discoveries when there is a vast change in microbial load or valid inference due to technical zeros. This study offers a valuable contribution to the field of microbiome studies by addressing limitations of existing methods and providing a new approach for analyzing compositional counts data with technical zeros.