Winnow-tax: sensitive and precise taxonomic profiling of low-coverage organisms in shotgun metagenomes
Accurate species-level profiling of shotgun metagenomes becomes difficult when an organism is represented by only sparse sequence coverage. Per-read k-mer classifiers can retain sensitivity under these conditions, but short reads from conserved or shared genomic regions can generate false-positive calls among closely related references. More conservative marker-gene and genome-containment…
The article discusses a novel approach to taxonomic profiling of low-coverage organisms in shotgun metagenomes called "winnow-tax." Traditional methods struggle to accurately identify species when an organism is represented by sparse sequence coverage. Sensitive per-read k-mer classifiers can maintain accuracy, but short reads from conserved regions can lead to false positives.
On the other hand, conservative marker-gene and genome-containment methods provide stronger specificity but may sacrifice sensitivity with sparse genomic sampling.
The winnow-tax pipeline addresses this trade-off by separating the initial candidate detection from genome-level confirmation. It uses Sylph, Kraken2, and a branch-rescue procedure to nominate candidate species. Then, reads are competitively recruited to a sample-specific reference set. The presence of a species is evaluated based on read support, observed genome breadth, and a Lander-Waterman-based breadth ratio.
In a controlled synthetic community study with 82 genomes and a human DNA background, winnow-tax demonstrated a stronger balance between precision and sensitivity compared to other methods like Kraken2/Bracken, Sylph, and MetaPhlAn 4. This advantage was most pronounced at low-coverage detection boundaries. In the CAMI III Toy Longitudinal Human Gut benchmark, winnow-tax showed higher sensitivity (0.803) than Sylph (0.672), albeit with lower precision (0.923 compared to 0.970) and a modestly higher F1 score (0.858 versus 0.793).
The method was also tested in a clinical enteric stool cohort with culture/PCR reference testing. Winnow-tax achieved the best overall performance for detecting Salmonella, identifying 37 out of 48 composite-positive samples with only one false-positive call. The results indicate that winnow-tax is an effective profiling strategy for weak taxonomic signals, retaining them during candidate generation while requiring solid genomic evidence in accordance with their sequencing depth before accepting species presence.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
