Unbiased probabilistic taxonomic classification for DNA barcoding.

Area of Science:

• Ecology
• Bioinformatics
• Genetics

Background:

• High-throughput sequencing of barcoding regions aids species identification in environmental samples.
• Existing methods struggle to translate confidence scores into unbiased taxonomic probabilities.

Purpose of the Study:

• To present PROTAX, a novel probabilistic method for DNA sequence taxonomic classification.
• To address limitations in current taxonomic assignment confidence scores.

Main Methods:

• PROTAX utilizes a statistical multinomial regression model.
• It incorporates reference sequences within a pre-defined taxonomic tree.
• PROTAX can use various similarity measures or outputs from other classifiers as predictors.

Main Results:

• PROTAX improves taxonomic classification accuracy compared to baseline TIPP and RDP classifiers.
• It successfully combines complementary data from BLAST and TIPP for robust classification.
• The method demonstrates accurate and unbiased classification even with up to 50% mislabeled reference sequences.

Conclusions:

• PROTAX offers a robust solution for unbiased taxonomic classification of DNA sequences.
• The method enhances the study of species diversity and community structure from environmental DNA.
• PROTAX accounts for unsequenced species, unknown taxa, and reference sequence errors.