Wastewater ARG Metagenomics · Preregistered Reproduction

The ermB Shortfall in Effluent

In three of nine effluent samples our ermB total is far below the published one, while the other four genes in the same samples agree. This page shows where the reads went and what the follow-up tests do and do not explain. The ermB work was exploratory and done after the scoring; it does not change the registered result.

0.22ermB ratio in E021 (165 of 759 reads)
0.71ermB ratio in E036 (548 of 773)
0.73ermB ratio in E011, a development sample (557 of 761)
6 of 9effluent samples agree within 4% on ermB

ermB, ours against published, per effluent sample

Ratio of our ermB total (templates with ≥ 50 reads) to the published ermB total. The shaded band is 0.90–1.10. Click a sample to see its templates below.

inside 0.90–1.10 outside E001, E006, E011 are development samples (Day 1); the others are evaluation samples

Where the reads went

For the selected sample: every template KMA reported for the erm(B) family (teal names) and the closest other erm genes (grey). Reads is KMA’s read-assignment count; coverage is how much of the template length is supported. A template with many assigned reads but low coverage is a “thin block”, not whole-gene evidence.

Sample
Template
Reads assigned (bar scaled within sample)
Template coverage (%)
coverage ≥ 95% (whole-gene support) coverage < 95% (thin block)

What the follow-up tests show

With competition (full database).
E021: erm(B)_2 receives 0 reads; the control E026 receives 698. Reads are assigned to erm(B)_18, erm(B)_20, erm(G)_1 or erm(F)_3 instead.
Without competition (erm(B)_2 alone, raw reads).
E021: 899 reads, 738 of 738 positions covered. E026: 906 reads, 738 of 738. The two samples look alike.

The single-template count is an upper bound: any sufficiently similar read, including from related erm genes, can map to it. It shows that sequence similar to erm(B)_2 is present, not how much of it there is.

HypothesisStatusEvidence
KMA reference competition / ambiguous allele assignmentsupportedReads exist in E021 at the control’s level when erm(B)_2 stands alone; in the full database they are labelled as neighbouring templates.
Assumed bbduk adapter optionsnot supportedRaw, untrimmed reads give the same absence in E021; trimming changes counts by about 2–17%.
Differences from the authors’ reference files or indexunresolvedCannot be tested without the authors’ exact files, which were not available.
Sample or run mix-up; error in the published tablenot supportedThe other four genes agree in the same samples; md5 checks passed. The shared published erm(B)_2 value of 659 in E011 and E021 is unexplained.

The same samples, all five genes

Ratio of our gene total to the published gene total, evaluation samples. Dashes mean the paper reports no count for that gene in that sample. Green is within 0.90–1.10; red is outside.

Gene-level agreement is strong everywhere except ermB in two samples. At subtype level, counts swap between near-identical alleles, which is the same effect seen here at larger scale.