A comparison of methods for clustering 16S rRNA sequences into OTUs

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• Next-generation sequencing of 16S rRNA revolutionized microbial community studies.
• Clustering 16S rRNA sequences into Operational Taxonomic Units (OTUs) is key for analyzing genetic diversity, composition, and richness.
• Existing OTU inference methods lack comprehensive performance guidance and parameter selection criteria.

Purpose of the Study:

• To comprehensively evaluate ten existing OTU inference methods.
• To determine the impact of dissimilarity thresholds and sequence abundance on OTU inference.
• To provide guidance on selecting appropriate parameters for OTU clustering.

Main Methods:

• Evaluation of ten different OTU inference algorithms.
• Analysis of 16S rRNA sequence data.
• Assessment of varying dissimilarity thresholds and sequence abundance parameters.

Main Results:

• OTU inference performance is dependent on both the chosen method and sample complexity.
• Low-complexity datasets require higher dissimilarity thresholds for OTU definition.
• Hierarchical clustering methods generally perform better at lower dissimilarity thresholds, but require stricter thresholds for high-complexity data.

Conclusions:

• The optimal dissimilarity threshold for OTU inference is method- and sample-dependent.
• Sequence abundance is a critical factor in accurate OTU inference.
• Careful selection of both dissimilarity and abundance thresholds is necessary for reliable microbial community analysis.