Data integrity in peptide QC records

A certificate of analysis is a summary written by the seller. The evidence behind it — instrument files, integration settings, sample weights, audit trails — is what makes the summary checkable. Most buyers never ask for it, which is precisely why asking changes the conversation.

Why a buyer should care

Data integrity is usually discussed as a regulatory topic for drug manufacturers. For a research peptide buyer it has a narrower, practical meaning: can the result on the certificate be traced back to a measurement that was made on this lot, and can an independent person reconstruct it? If the answer is no, the certificate is a statement of confidence rather than a record of evidence.

This is not an accusation of fraud. The common failures are ordinary: a template filled from the previous lot, a chromatogram saved as an image, a reintegration that was not documented, a mass spectrum reported without the raw file. Each is a gap in traceability, and each is fixable.

The principles, without the jargon

Principle	What it means on a peptide record
Attributable	The analyst and the reviewer are named, with dates and signatures; shared logins are not attribution
Legible	The record can be read without asking the author what an abbreviation meant
Contemporaneous	Recorded when the work was done, not reconstructed at shipment
Original	The raw instrument file or a true copy, not only a processed report
Accurate	Values match the raw data, including dilutions, weights and unit conversions
Complete	Failed injections, repeats, deviations and excluded data are visible, not deleted
Consistent	The same lot number, method and result appear across COA, batch record and label
Enduring and available	The record survives and can be produced on request within an agreed period

The list is long, but three items carry most of the weight in a purchase decision: original data, complete data, and attributable review.

Where peptide records commonly break

Area percent without integration detail. Purity by HPLC depends on how peaks were integrated. A chromatogram image with a percentage but no integration method, threshold or manual edits cannot be recalculated.
Mass spectrum as a picture. An annotated spectrum shows the intended mass; the raw file shows whether the peak was real, what else was in the sample and how the deconvolution was done.
No sample weight or dilution trail. A concentration or content result that cannot be traced to a weighed amount and a dilution scheme is not reconstructable.
Template values. The same appearance phrase, the same water result to one decimal, or a COA whose values are suspiciously round across lots suggests the template is doing the work.
Missing system suitability. Without the injection that demonstrates the method was working, the sample result has no context.
Deleted or excluded injections. Repeats are legitimate; repeats that vanish from the sequence table are not.
Shared accounts or disabled audit trails. If everyone signs as “lab”, the record cannot show who did what.
Dates that do not line up. Testing before sampling, review before acquisition, or a certificate dated before the batch record.

What to ask for

A request that is specific is more likely to be answered than “send raw data”. For a lot you are qualifying, ask for:

The instrument raw file for the chromatogram and, if run, the mass spectrum — not only the PDF report.
The acquisition and processing methods, including integration parameters and any manual integration events.
The sequence table: standards, blanks, system suitability, samples and repeats, in the order run.
The sample weight, dilution scheme and standard lot used for the calculation.
The audit trail or equivalent record of changes, if the system produces one.
The analyst and reviewer names, with dates.
The batch record summary linking the lot number to the material tested.

Not every supplier will have all seven. The pattern of what is available, and how quickly it is produced, is itself the assessment. A supplier who can send a raw file and a sequence table within a day is operating a system; one who can only send a new PDF is operating a template.

Reading an audit trail

An audit trail is a log of actions: who logged in, what was acquired, what was reprocessed, which peaks were manually adjusted, what was deleted. Three patterns are worth understanding:

Reprocessing is normal. Changing an integration parameter and reprocessing the same raw data is standard practice, provided the original remains and the change is recorded.
Manual integration is normal in moderation. A baseline adjustment on a tailing peak is routine. A series of manual edits that turns a shoulder into a single peak is a question.
Deletion is the line. Removing an injection or a result without a documented reason is not reprocessing; it is loss of the record.

You do not need the instrument software to read these patterns. A change log with timestamps, user names and before/after values is enough to see whether the record was controlled.

Paper, PDF and electronic records

A paper printout can be a valid record if it is complete, signed and contemporaneous. A PDF can be a true copy or a re-typed summary; the format alone does not tell you which. The useful questions are about completeness and linkage, not about paper versus electronic:

Does the printout include the method, the sequence and the system suitability, or only the result?
Is the file name or record ID traceable to the lot?
Can the supplier produce the underlying file if asked?
Is there a second-person review, and is it dated after the analysis?

For a small supplier without a laboratory information system, a signed paper record with the raw file attached is perfectly adequate. The absence of software is not the problem; the absence of linkage is.

What this cannot prove

Even a complete record does not prove that the file belongs to the lot you received. A determined seller can present a consistent package for different material. Data integrity reduces the chance of ordinary error and makes fabrication harder; it does not replace an independent check. That is the role of a third-party sample, sent blind and tested by a laboratory you chose. The two controls are complementary: records tell you what the supplier did, and independent testing tells you what the material is.

Writing it into the agreement

Three clauses make the request routine rather than a favour:

Records retention: batch and analytical records kept for an agreed period, and produced on request.
Raw data access: instrument files and processing methods available for any lot, within an agreed time.
Notification: the buyer is told of any OOS result, deviation, reprocessing or change affecting the lot.

These belong with the other technical terms in the specification annex and the procurement contract, not in a separate quality side-letter that nobody reads at renewal.

For laboratory research use only. This article describes record-keeping practice; it makes no claim about the records of any specific supplier and does not support administration to humans or animals.

Editor's take

A record you cannot trust is worse than no record, because it looks like evidence. Audit trails, version control and signatures are what separate data from decoration.

Limitations

This guide describes methods and interpretation, not acceptance limits. Limits must come from your own specification and, where applicable, the relevant pharmacopoeial monograph. No testing was performed by the author for this page.