Derived from 11,772 genomes: 299 from RefSeq and 11,473 from other sources. For the derivation pipeline and the PASS / WARN / FAIL verdict model, see the methods page for qualibact-v1.1.
Applied to the full All-The-Bacteria dataset, these thresholds place 10,503 genomes at PASS, 1,422 at WARN, and 575 at FAIL (12,500 assessed in total). The per-tier genome lists can be downloaded below in .csv.gz format; the FAIL list also records the reason each assembly was rejected.
QualiBact qualibact-v1.1 thresholds for Haemophilus influenzae, refined from v1.0 based on the Nov 2025 expert-feedback survey.
Two pins on her recommendation:
All other thresholds inherit unchanged from v1.0. WARN tier preserves the engine's tighter values. See the methods page for qualibact-v1.1 for the full pipeline.
Threshold values and rationale for Haemophilus influenzae (qualibact-v1.1) contributed by:
Nov 2025 expert-feedback survey — flagged that >100 contigs are most often contamination, and that two RefSeq references pushing the assembly-size upper bound up have inconclusive ANI (~94 %) against H. influenzae
This table summarises the distribution of each metric, including standard deviation, mean, median, and percentiles.
A combined summary table across all species is available on the summary page.
| Metric | Distribution | n | Mean | SD | Min | Q1 | Median | Q3 | Max |
|---|---|---|---|---|---|---|---|---|---|
| N50 | non-normal | 11,473 | 284,724 | 268,108 | 16,210 | 122,077 | 202,427 | 322,198 | 1,158,059 |
| no_of_contigs | non-normal | 11,473 | 43.2 | 19.61 | 12 | 30 | 40 | 52 | 223 |
| longest | non-normal | 11,473 | 476,129 | 246,470 | 56,596 | 287,119 | 502,311 | 563,401 | 1,158,059 |
| GC_Content | non-normal | 11,473 | 37.99 | 0.1 | 37.73 | 37.92 | 37.99 | 38.05 | 38.38 |
| Completeness_Specific | non-normal | 11,473 | 100 | 0 | 99.98 | 100 | 100 | 100 | 100 |
| Contamination | non-normal | 11,473 | 0.16 | 0.28 | 0 | 0 | 0.04 | 0.18 | 3.76 |
| Total_Coding_Sequences | non-normal | 11,473 | 1,783 | 70.25 | 1,642 | 1,729 | 1,774 | 1,830 | 2,108 |
| Genome_Size | non-normal | 11,473 | 1,848,417 | 50,139 | 1,743,262 | 1,810,454 | 1,842,859 | 1,881,744 | 2,104,737 |
Full statistics including KS test vs RefSeq and Wasserstein distance are in the downloadable summary.csv.
Derived from 11,772 genomes including 299 RefSeq references
Both Fail and Warn bands shown as the published rounded values — easier to cite and consistent across the species page, CSV downloads, and downstream QC tools.
| Metric | Fail below | Warn below | Warn above | Fail above |
|---|---|---|---|---|
| Genome_Size | 1,700,000 | 1,700,000 | 2,000,000 | 2,000,000 |
| GC_Content | 37.7 | 37.8 | 38.3 | 38.5 |
| Total_Coding_Sequences | 1,600 | 1,600 | 2,000 | 2,200 |
| Completeness_Specific | 93 | 100 | - | - |
| Contamination | - | - | 1 | 6 |
| N50 | 30,000 | 52,000 | - | - |
| no_of_contigs | - | - | 100 | 100 |
| longest | - | - | - | - |
How to read this: a value between the two warn columns is typical for this species and passes QC. A value between a warn column and the corresponding fail column is borderline — worth a manual look but not an outright failure. A value outside the fail columns is unusual enough to fail QC.
The published rounded thresholds (the values in the table above) were applied to the full AllTheBacteria-2024-08 set for this species. Each row carries the per-metric verdict and, where applicable, the reason a genome was demoted to WARN or FAIL. Files are gzipped CSV.
This plot shows the relationship between the number of coding sequences (CDS) and genome size — how the number of genes scales with assembly length. The relationship should be roughly linear: as genome size increases, the number of coding sequences should rise proportionally. A secondary trend line or non-linear behaviour can indicate either bona fide sub-populations within the retained genomes (e.g. distinct sub-clades) or residual contamination that survived filtering.
Histogram comparing SRA to RefSeq; each bar shows genome density across value ranges to highlight shifts, peaks, or outliers.
QQ (quantile-quantile) plot comparing SRA and RefSeq. Points along the diagonal follow the expected distribution; deviations indicate skew, outliers, or other systematic differences.
A table of complete RefSeq genomes for Haemophilus influenzae used to calibrate this scheme. The file includes accessions, some sample information, genome size, GC content, and other key metrics.
These plots show genomes before and after filtering to highlight the outliers removed:
The filtered distribution shown here may not exactly match the published thresholds because additional rounding and curator adjustments are applied on top.
