When Nature surveyed 1,576 researchers on reproducibility, more than 70 per cent said they had tried and failed to reproduce another scientist’s experiment and more than half had failed to reproduce their own. The discussion that followed concentrated on study design, statistics and selective reporting. It spent much less time on the layer underneath, where a number that looks like a measurement is partly a processing decision.
Chromatographic purity is one of those numbers. The detector records a signal. Software converts that signal into peak areas. The conversion is governed by parameters an analyst sets, and two analysts processing the same raw file with different parameters can report different purity figures for the same material without either of them doing anything improper.
The decisions hiding inside a percentage
Several processing choices affect the result, and most certificates do not record any of them.
Baseline placement is the largest. Where the baseline is drawn beneath a cluster of peaks determines how much area is assigned to each. Drawing it across the valley between two partly overlapping peaks gives a different split than dropping a perpendicular from the valley floor, and the difference is largest exactly where it matters most, in the region immediately around the product peak where close relatives elute.
Threshold and minimum area settings decide what counts as a peak at all. Raise the threshold and small signals are classified as noise and excluded from the impurity table. Nothing has been concealed; the setting simply defined them out of the calculation. Lower it and baseline wander begins generating peaks that have no chemical meaning.
Peak width and slope sensitivity govern how the software recognises the start and end of a peak. Set for sharp early peaks, they can clip the tail of a broad late one. Set for broad peaks, they can merge two narrow neighbours into one.
And tangential skim versus perpendicular drop, applied to a small peak riding on the shoulder of a large one, is a straightforward fork in the road: one approach assigns most of the shared area to the large peak, the other distributes it differently. For a related substance sitting on the shoulder of the product, that choice alone can move the reported impurity level appreciably.
Why area percent is a convention, not a measurement of content
There is a second assumption built into every purity figure expressed as area percent, and it is a bigger one than the integration settings.
Area percent treats the proportion of total detector response as equivalent to the proportion of material present. That holds only if every component responds to the detector in proportion to how much of it there is, at the same rate. For ultraviolet detection the response depends on the chromophores present, so a related substance that has lost an absorbing residue is under-represented relative to its true abundance, and one that has gained one is over-represented.
This is why chromatographic purity and peptide content are different quantities answering different questions, and why amino acid analysis exists as a separate determination. A certificate reporting a high area-percent purity has described the distribution of detector response. It has not weighed anything.
The case where it bites hardest
Integration choices matter most when peaks are poorly resolved, which for synthetic peptides is the normal condition rather than the exception.
Process-related impurities from solid-phase synthesis are structural relatives of the target: single-residue deletions, insertions, truncations, incompletely deprotected intermediates. ICH Q2(R2) identifies the resolution of the closest-eluting pair as the way specificity is demonstrated, with a working criterion above 1.5, and published method guidance names co-elution of deletion sequences as the central specificity challenge in peptide analysis. A pair sitting just above that criterion is a pair where baseline placement and skim choice carry real weight.
Short, heavily modified sequences concentrate the problem further. Terminal modifications and non-standard residues produce related substances whose retention and response are harder to predict, so the analyst has less basis for judging whether a given integration approach is the sensible one.
How it works in practice
A purchaser cannot audit someone else’s data processing, and does not need to. What they can do is check whether the document admits that processing happened.
The chromatogram is the disclosure. A trace shows where the baseline was drawn, whether peaks were resolved or merged, and whether the region around the product is clean or crowded. It lets a reader form an independent view of how much the reported figure depended on judgement. A purity percentage presented alone does not.
For a short modified sequence such as ss-31 supplied for laboratory research use, where related substances are close structural relatives and resolution is tight, the attached trace carries most of the usable information in the certificate. For laboratory research use only. Not for human or veterinary use. The questions that follow from it are practical: is the region around the product peak resolved or crowded, is the reported figure area percent or a content determination, and is the trace specific to the lot in hand rather than a representative example from the product line.
Suppliers that attach lot-specific traces and state which quantity they are reporting make those questions answerable. That is the whole of what a purchaser can reasonably expect, and it is enough.
Where the convention is moving
Regulated laboratories have tightened processing governance considerably, treating integration parameters as part of the method rather than as analyst discretion, with manual reintegration requiring justification and an audit trail. The logic is that a reported number should be reproducible from the raw data by someone else, which is only true if the processing rules travelled with it.
Research-grade supply has adopted the visible half, attaching traces, and has not adopted the other half, publishing processing parameters. That is unlikely to change, since no external requirement applies to research-use-only material. It does mean the trace is doing double duty: it is both the evidence for the number and the only window onto how the number was produced.
Conclusion
A purity percentage is a processed result. Baseline placement, detection thresholds, peak width settings and the handling of shoulder peaks all move it, and area percent additionally assumes a uniform detector response that the chemistry does not guarantee.
None of that makes the figure useless. It makes it a summary that should be read next to the thing it summarises. Ask for the trace, look at the region around the product peak, and treat a purity number offered without one as a claim rather than a measurement. The difference between those two is the difference between a result another laboratory could reconstruct and a result it can only accept.
