Provenance — who produced it, who reused it
Linked to 1 papers in the literature. Roles are inferred factual signals (who deposited the data vs who reused it), with counts — never a judgement about any author.
1 further paper cites this accession but reuse could not be confirmed.
Deep data QC
56/100 · FStandardized, field-standard QC computed by touching the data — every metric states how it was obtained · evidence: measured
Whole-genome shotgun sequencing of a Lactobacillus (Lacticaseibacillus zeae) isolate on Illumina MiSeq, with the long 301 bp reads characteristic of that platform. Overall this dataset fails QC (56/100, grade F), and the verdict is dominated by one severe problem: an adapter content of 54.47% means more than half the reads carry residual adapter sequence, which is what zeroed out that metric and would corrupt assembly and mapping unless the reads are aggressively trimmed before any reuse. The base quality is mediocre but usable — Q30 at 82.1% and a mean base quality of 33.5 sit in acceptable-but-unremarkable territory — while the low 10.08% duplication rate and 0.024% N content are genuine strengths, so the data are salvageable if you trim hard rather than use them as-is. One practical caveat for the WGS use case: with only ~137 Mb of total bases the depth over even a small Lactobacillus genome is modest, so variant-calling sensitivity will be limited; note also that the core size/read-count figures are reported rather than independently measured, so confirm the post-trimming yield yourself before committing to this dataset.
The F grade is a transparent weighted average. Each metric below scored from 0–100% against the published WGS thresholds, weighted by its importance; nothing is hidden or subjective.
measured = computed from the data · extrapolated/reported = derived or from the repository · dq-1.0
Scientific quality
Based on hands-on reproduction of the papers that use this dataset. A reproducible paper that stands on this data is positive evidence; a flagged one is a prompt to look closer — never a verdict on the dataset itself without the evidence.