Taxonova: Deep learning enables fast and confidence-aware microbiome species-level amplicon classification.

Taxonova enables species-level classification of ribosomal amplicons across 250 bp–4,900 bp, achieving median F-measures of 0.94 (short-read) and 0.97 (long-read).

• Why it matters: Species-level resolution is hindered by conserved ribosomal regions, short amplicon length, and uneven database coverage, leaving no robust method across amplicon lengths and sequencing targets.
• What they did: The deep learning framework was evaluated on physical and biome-simulated mock communities spanning seven biomes using Taxobench, and benchmarked against other classifiers for accuracy, abundance error, and false positives.
• The result: Taxonova reduces species-level abundance-estimation error (L1 norm) by a median of 0.52–1.50 and lowers species-level false-positive rate by a median of 16.2 × 10-4, while matching shotgun profiles with higher Jaccard concordance.

AI summary of the abstract · may contain errors · not medical advice · read the paper.