How to actually read a Certificate of Analysis: a field guide for skeptical researchers

Every vendor in this space will hand you a Certificate of Analysis. A PDF with a chromatogram, a purity number with two decimal places, maybe a mass-spec trace. The document looks authoritative. That is precisely the problem — looking authoritative is cheap, and most researchers have never been taught what to actually check.

I have audited hundreds of COAs, ours and competitors'. This is the checklist I use, in the order I use it. None of it requires special software — just the PDF and ten minutes of attention.

1. Start with the lot number, not the purity

The purity figure is the last thing to look at, because it is the easiest to fabricate. Start with provenance:

• Does the lot number on the COA match the vial label? Character for character. A COA without a lot number is marketing material, not a certificate.
• Who issued it? A COA generated by the vendor's own bench is a self-report card. Look for the testing laboratory's name, address, and accreditation. Independent issuance is the single strongest signal.
• Dates. Synthesis date, test date, and issue date should be present and in a plausible order. A test date six months before the synthesis date happens more often than you would think — it means the COA was recycled from an earlier lot.

A recycled COA is not a shortcut, it is a forgery. If a vendor reuses one document across lots, nothing else on the page means anything.

2. Read the chromatogram like a skeptic

The HPLC trace is where most games are played. Four things to check:

Retention time is stated and plausible

The main peak should have a labeled retention time (RT). For a given column, gradient, and compound, RT is reproducible within a narrow window. If the COA does not state the method (column, mobile phase, gradient, wavelength), the RT is uninterpretable — and that omission is a choice.

The baseline is visible

A chromatogram cropped tight around the main peak hides the small stuff — the shoulder peaks and late eluters where the real impurity story lives. You want to see the full run, injection artifact to column wash.

Integration windows make sense

Purity-by-HPLC is the main peak's area divided by total integrated area. If the integration ignores a visible shoulder, the stated purity is inflated. Look for the peak table: every visible peak should appear in it.

The y-axis scale is honest

At full scale, every trace looks clean. Reputable labs include a zoomed inset (or a second plot) at a scale where 0.1% impurities are visible.

3. Mass spec: the identity check

HPLC tells you the sample is pure; it does not tell you it is the right molecule. A clean chromatogram of the wrong peptide is still the wrong peptide. The COA should report:

For larger peptides you will often see multiply-charged species (M+2H)²⁺, (M+3H)³⁺ rather than the naked molecular ion. That is normal electrospray behavior, not a discrepancy — but the deconvoluted mass should still land on theory.

4. The tests that are usually missing

Purity and identity get all the attention, but three quieter numbers separate a real QC program from a printed one:

1. Water content (Karl Fischer). Lyophilized peptides are hygroscopic; 5–8% residual water is common and means your "10 mg" vial holds less peptide than labeled. We spec ≤6% and report the actual figure.
2. Counter-ion content. Peptides purified with TFA carry trifluoroacetate salts — often 10–20% by weight. If the COA is silent on TFA, the net peptide mass is unknown.
3. Endotoxin (LAL). Matters for any cell-based work. "<0.5 EU/mg" is a common spec; we hold to <0.05. If endotoxin is not tested, cell-culture researchers should assume the worst.

5. A 60-second checklist

Print this, tape it to the fridge in the lab:

• ☐ Lot number matches the vial
• ☐ Independent lab named and accredited
• ☐ Dates present and ordered sanely
• ☐ HPLC method stated; full-window trace; peak table complete
• ☐ MS observed mass within tolerance of theoretical
• ☐ Water content and counter-ion reported
• ☐ Endotoxin tested if cell work is planned

Any single missing item is a question to send the vendor. Two or more is an answer.