A Bayesian taxonomic classification method for 16S rRNA gene sequences with improved species-level accuracy

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Species-level classification of 16S rRNA gene sequences presents a significant challenge in microbiome research.
• Existing tools often lack reliable species-level classification or accurate confidence metrics, hindering research.
• Current methods rely on limited criteria like k-mer frequency, failing to capture true sequence similarity.

Purpose of the Study:

• To develop a novel method for accurate and reliable species-level taxonomic classification of 16S rRNA gene sequences.
• To improve upon the limitations of existing classification tools by incorporating probabilistic criteria.
• To provide researchers with a robust tool for microbiome data analysis.

Main Methods:

• Developed a method utilizing pairwise sequence alignment to calculate true sequence similarity.
• Employs a Bayesian approach, weighting database hits by posterior probability based on sequence similarity.
• Assigns taxonomy from species to phylum levels using lowest common ancestors and evaluates reliability with bootstrap confidence scores.

Main Results:

• Achieved significantly improved species-level classification accuracy compared to existing methods.
• The method provides probabilistic confidence scores for taxonomic assignments.
• Demonstrated applicability across different 16S rRNA gene regions and other phylogenetic markers without specific training datasets.

Conclusions:

• The developed software, BLCA, offers reliable species-level classification for 16S rRNA and other phylogenetic marker genes.
• BLCA provides accurate taxonomic assignments with confidence scores, enhancing microbiome research reliability.
• Despite higher computational demands, BLCA is suitable for large-scale microbiome datasets and is freely available.