How a Peptide Specification Limit Gets Set
“Purity: not less than 98%” looks like a specification and is not one. Purity measured how, by chromatographic area, by mass balance, by net peptide content? Read at which wavelength, across which column, over which gradient? Strip the method away and what remains is an assertion, the same weakness that afflicts grade names in crude, desalted and purified grades. On a certificate, the value beside a limit is a measurement; the limit itself is somebody’s decision, and how that decision got made is what determines whether passing it means anything.
Three parts, all required
Anything worth calling a specification names an attribute, names a method, and states an acceptance criterion. Remove one of those three and it ceases to be a specification at all. Attribute and method jointly establish what is under constraint. The criterion marks where the boundary falls.
Can the method decide the limit?
Tighten a limit past what the method can resolve and the limit becomes ornamental. Suppose chromatography quantifies dependably down to 0.1%. A criterion reading “any single impurity not more than 0.05%” cannot be evaluated with it, and recording a pass against that criterion records a measurement nobody performed. Those thresholds are laid out in limit of detection and limit of quantitation.
The failure also runs in the other direction. Assess “not less than 98.0%” using a method carrying an uncertainty of plus or minus one percent and you have written a limit the method has no power to adjudicate, which is the subject of measurement uncertainty. Choosing a limit and validating a method are one activity, not two.
Four places a number can come from
Limits have four possible origins, carrying rather different weight.
| Compendial monograph | Where a pharmacopoeial entry exists, both limit and method are laid down for you. Few research peptides have such an entry, which shuts this route for most of them. |
| Safety or toxicological basis | Applied to impurities where a quantity carries known significance. Elemental impurities and residual solvents are the familiar cases, taken up in elemental impurities and residual solvents. |
| Process capability | The figure reflects what manufacturing achieves consistently, usually observed lot history plus a margin. This is the commonest real basis, and it describes a process rather than fitness for any purpose. |
| Commercial convention | Ninety-eight percent is what the market expects, not something derived. A limit arrived at this way amounts to a promise, which is honest enough provided nobody pretends it was calculated. |
In this market the fourth row outnumbers the first three by a wide margin. That is not objectionable by itself, since a conventional limit met consistently still carries information, but it deserves to be read for what it is.
What a specification passes over in silence
Constraint extends only to the attributes actually named. Everything unnamed floats free, no matter how many decimal places decorate the attributes that were named.
Consider a peptide comfortably satisfying a 98% chromatographic purity criterion while carrying 25% counter-ion by weight and 8% water. Neither figure is an attribute of that specification, so neither is constrained by it. That is precisely the distinction developed in net peptide content, and it represents a hole in the specification rather than a defect in the material. Reading well means registering which attributes are missing, not merely scanning the values present.
One attribute, two criteria
Careful specifications frequently give an attribute two limits: a stricter one applying at release, a looser one covering the remainder of the assigned period.
Degradation being predictable, a lot released right at the shelf-life limit would drop below it within a month. Stability data of the kind described in what a stability protocol contains is what sets the distance between the two, and it explains why an assigned date carries different meaning in different systems, as covered in retest dates and expiry dates.
Where a single criterion serves both moments, the specification is either unduly strict at release or unduly hopeful at the far end. Experience suggests the latter.
Three phrasings, three different amounts of information
Certificates use “not detected”, “not more than X” and “complies” as though the three were interchangeable, and they are not. The first reports that nothing showed above the detection threshold of whichever method ran, which says as much about the method as about the sample. The second reports that a measurement happened and came in under X. The third reports only that some criterion was evaluated, withholding the value, which is exactly the piece a buyer would need to set two lots side by side. Wherever the number exists, printing it costs the issuer nothing and gives the reader more than a verdict ever will.
Interrogating a specification you did not write
The useful questions are few. Which attributes carry constraints, and which obvious ones escape them? Does every criterion arrive paired with a method? Could that method plausibly decide that criterion? Are you given a value or only a judgment? Was the limit derived, or is it conventional?
Answering all five takes no laboratory, and together they sort reports that convey information from reports that convey comfort, which is much the same line that divides the document types compared in certificate of analysis, conformance and origin. Wider framing sits in how to read a certificate of analysis, and the reasons two scrupulous certificates for one material can still fail to agree are in why certificates disagree on purity.
