HiFiTaxa: a reproducible analysis pipeline for full length 16S rRNA gene and ITS amplicon sequencing from PacBio reads

HiFiTaxa shows that for PacBio HiFi 16S rRNA gene mock communities of 20 species, BLCA detects all 20 with two false positives, while NB detects 17 with none.

• Why it matters: Taxonomic resolution from long-read amplicon sequencing is limited by how references, classifiers, and parameters are chosen, so reproducible, benchmarked workflows are needed.
• What they did: The single-command Nextflow pipeline builds versioned GTDB and UNITE references and supports BLCA and naive Bayes classifiers plus Emu (16S) and EMITS (ITS) read-level profilers, then benchmarks on mock communities and holdouts with 10% withheld.
• The result: In GTDB holdouts, BLCA reaches 80.0% species-level accuracy versus 48.1% for NB’s exact matching step, while NB yields fewer incorrect assignments for taxa absent from the reference.

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