raxtax: a k-mer-based non-Bayesian taxonomic classifier

Area of Science:

• Biodiversity studies
• Bioinformatics
• Genomics

Background:

• Accurate taxonomic classification is crucial for biodiversity studies, enabling the assessment of biological system diversity.
• Current methods face challenges in accuracy and speed, especially with growing reference databases.
• Errors in classification can significantly impact downstream analysis results.

Purpose of the Study:

• To introduce raxtax, an efficient and novel tool for taxonomic classification of barcodes.
• To address the accuracy and speed limitations of existing taxonomic classification tools.
• To develop novel uncertainty scores that account for reference database biases.

Main Methods:

• raxtax utilizes common k-mers between query and reference sequences for classification.
• The tool incorporates two novel uncertainty scores to address reference database biases.
• Validation was performed on three empirical reference databases.

Main Results:

• raxtax demonstrates significant speedups, being 2.7-100 times faster than state-of-the-art tools on large databases.
• The tool maintains accuracy comparable to existing methods.
• Increasing query and reference sequence numbers result in greater speedups, alleviating scalability challenges.

Conclusions:

• raxtax offers an efficient and accurate solution for taxonomic classification of barcodes.
• The tool's performance improves with larger datasets, addressing scalability issues in biodiversity research.
• Novel uncertainty scores enhance the reliability of classification by considering database biases.