Microbiome Datasets Are Compositional: And This Is Not Optional.

Area of Science:

• Microbiology
• Bioinformatics
• Data Science

Background:

• High-throughput sequencing (HTS) generates microbiome datasets (16S rRNA, metagenomes, metatranscriptomes) widely used in human disease, ecology, and environmental studies.
• These HTS datasets are inherently compositional due to arbitrary instrument-imposed totals, a fact often overlooked by researchers.
• Misunderstanding or ignoring the compositional nature of microbiome data can lead to analytical errors and flawed conclusions.

Purpose of the Study:

• To highlight the critical importance of recognizing and addressing the compositional nature of HTS microbiome data.
• To warn researchers about the potential pitfalls of inappropriate data analysis when compositional properties are ignored.
• To provide guidance and resources for applying compositional data analysis (CoDA) techniques to microbiome studies.

Main Methods:

• Review of existing literature on microbiome data generation and analysis.
• Introduction to the principles of compositional data.
• Illustration of analytical issues arising from non-compositional approaches.
• Guidance on applying compositional data analysis (CoDA) methods.

Main Results:

• Demonstration of how treating compositional data as standard data can lead to "pathologies" or incorrect results.
• Emphasis on the necessity of using CoDA methods throughout the analysis pipeline for HTS microbiome datasets.
• Identification of resources and examples for implementing CoDA in microbiome research.

Conclusions:

• Microbiome datasets from HTS are compositional and require specialized analytical approaches.
• Adoption of compositional data analysis is crucial for accurate interpretation of microbiome studies.
• Researchers are encouraged to utilize CoDA to avoid analytical errors and advance microbiome research.