1,476 published peptide laboratory reports across 223 vendors, with documentation-quality and publication-selection fields. Open data, CC BY 4.0, with the analysis script that generates every published figure.
This page is the citable source for the figures in The COA Mirage. It exists so that anyone who wants to check a number, re-run the analysis, or use the corpus for their own work can do so without asking us first.
The corpus is 1,476 published analytical reports attributed to Janoshik Analytical, across 223 vendor identities, spanning February 2020 to 6 May 2026 plus 22 undated reports. As far as we know it is the largest assembled set of published peptide certificates anywhere.
Everything below is free to use under CC BY 4.0. We accept no vendor sponsorship, sell no testing, and take no affiliate revenue from any company named in the data.
| File | Contents |
|---|---|
coa-corpus-v1.csv | The full corpus. One row per published report; eight columns. 1,476 rows. |
coa-corpus-v1-codebook.md | Column definitions, selection model, known limits, reproduction instructions, update log. |
coa-corpus-stats.json | Every published figure, as emitted by the analysis script. |
figures/ | The three charts in the article, as PNG. |
46.7% of published certificates carry no readable purity figure. Of 1,476 reports, 690 contain no numeric purity value at all — while 1,470 of them do carry a certificate image. These are largely genuine analytical documents: identity confirmations, mass-spectrometry reports without quantitation, cropped scans. They look like proof of purity and are not evidence for it.
The 99.4% pass rate measures publication behaviour, not product quality. Among the 786 reports that do carry a readable value, 99.4% are at or above 95%, with a median of 99.7%. Every certificate here crossed a publication barrier: someone ordered a test, received a result, and chose to make it visible. Failed assays that were never posted, batches retested until one passed, and batches never tested at all are invisible to this dataset by construction.
If you take one sentence from this page, take that second one. It is the finding, and the near-perfect pass rate is the evidence for it rather than a contradiction of it.
The article promises that the script generating its figures is public. It is:
node node_modules/tsx/dist/cli.mjs scripts/research/coa-corpus-stats.ts
It reads coa-corpus-v1.csv, writes coa-corpus-stats.json, and the article renders from that JSON. No published figure is hand-typed, so the page and the data cannot drift apart.
SHA-256 of coa-corpus-v1.csv:
c7957abfd9eb73ea95011eaae43750e9075bb723ea209af1e1437329b6875c03
0 invents 690 failures. Filter, do not fill.The full statement of limits is in the codebook, and it is longer than this list.
v1 is frozen and will not change. The live corpus behind this site is rebuilt continuously by our pipelines, and it does not only grow — it has lost rows to pipeline failure before. A dataset whose row count drops after someone cites it is worse than no dataset. New records become v2, published alongside v1 at its own URL and hash, never replacing it.
| Date | Version | Change |
|---|---|---|
| 2026-08-26 | v1 | Initial public release. 1,476 reports, 223 vendors, frozen. |
We would rather give you the data than a quote.
The corpus, the codebook and the analysis script are above with no gate on them. If you need something that is not here — a cut we have not published, the underlying certificate images, the correspondence behind the supply-chain evidence, or a sanity check on a number before you print it — write to [email protected] and we will answer, including when the answer is that we do not have it.
We are not asking for a link. If you use the data, cite the dataset in whatever form your publication uses.
> PeptideAtlas. *Grey-Market Peptide COA Corpus v1* [data set]. August 2026.
> https://peptideatlas.co/research/grey-market-peptide-coa-dataset
A DOI deposit is in progress and will be added to this page and to the codebook when it is issued.
Related: The COA Mirage.