Taxonomy annotation and guide tree errors in 16S rRNA databases.

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• 16S ribosomal RNA (rRNA) gene sequencing is a cornerstone for microbial community analysis.
• Existing databases like Greengenes, RDP, and SILVA provide taxonomic annotations and guide trees for this approach.
• Current annotations often rely on sequence prediction rather than authoritative type strain data.

Purpose of the Study:

• To critically evaluate the accuracy of taxonomy annotations and guide trees in major 16S rRNA gene databases.
• To identify and quantify discrepancies between database annotations, guide trees, and established taxonomic references.
• To assess the reliability of these databases for phylogenetic inference in microbial ecology.

Main Methods:

• A blinded test was employed to estimate the annotation error rate of the RDP database.
• Branching orders of Greengenes and SILVA guide trees were compared against each other and the RDP training set.
• Identical sequences from SILVA and Greengenes databases were analyzed for annotation conflicts across various taxonomic ranks.

Main Results:

• The RDP database exhibited an estimated annotation error rate of approximately 10%.
• Greengenes and SILVA guide trees showed comparable disagreement rates with each other and with RDP's authoritative taxonomy.
• Over 249,000 identical sequences displayed conflicting annotations between SILVA v128 and Greengenes v13.5, indicating a ~17% annotation error rate.

Conclusions:

• The guide trees in Greengenes and SILVA are unreliable for inferring microbial phylogeny due to pervasive conflicts with type strain taxonomies.
• Significant annotation errors exist across major 16S rRNA gene databases, compromising the accuracy of microbial community surveys.
• The findings underscore the urgent need for improved database curation and validation to ensure the reliability of microbial research.