Why two labs report different purity for the same peptide lot: RP-HPLC method variables

Send a single research peptide lot to two independent analytical laboratories and it is entirely ordinary to receive back two different purity figures — 98.2% from one, 99.1% from the other. Neither laboratory has made an error. Both numbers are defensible outputs of the methods that produced them. The discrepancy arises because “purity by HPLC” is not a physical constant of the material; it is the result of a measurement whose value depends on how the measurement was configured.

Reversed-phase high-performance liquid chromatography (RP-HPLC) is the dominant purity method for synthetic peptides, and the number it produces appears near the top of nearly every certificate of analysis. Understanding what that number is sensitive to is the difference between reading a COA as a specification and reading it as a measurement with a method attached. This post covers the principal method variables — stationary phase, gradient, detection wavelength, and integration practice — and what each one does to the reported figure.

What the chromatogram is actually measuring

RP-HPLC separates a mixture by partitioning analytes between a hydrophobic stationary phase, typically octadecylsilane (C18) bonded to silica, and a polar mobile phase that becomes progressively less polar over the run. A peptide adsorbs to the stationary phase at the starting mobile-phase composition and elutes when the organic fraction rises enough to overcome that interaction. Compounds that differ in hydrophobicity elute at different times, and a detector at the column outlet records each as a peak.

The purity figure is then calculated by area normalization: the area of the main peak divided by the summed area of all integrated peaks, expressed as a percentage. This calculation carries an assumption that deserves to be made explicit — it assumes every species in the sample produces detector signal in proportion to its mass. That assumption is approximately true for closely related peptide impurities and substantially false for anything structurally dissimilar. Area normalization is a relative measure of the chromatographic profile, not an absolute assay of how much peptide is present. That second question is answered by peptide content determination, usually by amino acid analysis or nitrogen determination, and it is a separate line on the COA.

Stationary phase and column geometry

Column selection is the single largest source of between-laboratory variation. A C18 phase is standard, but C18 columns from different manufacturers differ in pore size, particle diameter, carbon load, endcapping chemistry, and residual silanol activity. For peptides above roughly 3 kDa, pore size matters considerably: a 300 Å wide-pore packing allows the molecule to access the internal surface area of the particle, whereas a 100 Å packing largely excludes it, compressing retention and degrading resolution between the main peak and its nearest impurities.

Particle diameter and column length together set the plate count, and plate count sets how well two closely eluting species are resolved. A sub-2-micron UHPLC column can baseline-resolve a deamidation product that a conventional 5-micron column reports as a shoulder on the main peak. When that shoulder goes unresolved, its area is integrated into the main peak, and the reported purity rises. The material has not changed. The resolution has.

This is why a purity figure without a stated method is difficult to interpret comparatively. A well-constructed COA names the column, its dimensions, and the particle and pore specification. Its absence does not indicate a problem, but its presence allows a reader to judge whether two figures are being compared on equivalent footing.

Gradient slope and run time

The mobile phase in peptide RP-HPLC is typically water and acetonitrile, both containing an ion-pairing acid — most commonly 0.1% trifluoroacetic acid (TFA). The organic fraction is increased over the run according to a programmed gradient, and the slope of that gradient trades resolution against analysis time.

A shallow gradient — for instance 1% acetonitrile per minute across a narrow window bracketing the elution point — spreads closely related species apart and reveals impurities that a steep gradient co-elutes with the main peak. A steep gradient of 3–5% per minute completes the run quickly and, predictably, reports a higher purity number, because unresolved material is being counted as product.

The ion-pairing agent contributes its own effects. TFA forms ion pairs with basic residues, masking their charge and sharpening peaks considerably. Substituting formic acid, which is often done when the method is coupled to mass spectrometry because TFA suppresses electrospray ionization, generally produces broader peaks and lower resolution. A laboratory running an LC-MS-compatible method and a laboratory running a dedicated UV purity method are therefore not running the same separation, and their integration outcomes can differ systematically.

Detection wavelength and the response-factor problem

Peptide UV detection is performed at either 214 nm or 280 nm, and the choice materially changes what the chromatogram shows.

At 214 nm the detector responds primarily to the amide bond of the peptide backbone. Because every residue contributes a backbone carbonyl, absorbance at this wavelength scales roughly with peptide length and is relatively uniform across species. This makes 214 nm the general-purpose choice for purity work and the wavelength most commonly specified on peptide COAs.

At 280 nm the detector responds to aromatic side chains — tryptophan, tyrosine, and to a much smaller degree phenylalanine and cystine. A peptide containing none of these residues is nearly invisible at 280 nm. The wavelength is useful for concentration determination in peptides with known aromatic content, and it can be diagnostic when a laboratory wants to distinguish an aromatic-containing impurity from a non-aromatic one, but it is a poor general purity wavelength.

The response-factor issue persists even at 214 nm. A truncated synthesis byproduct missing several residues has fewer amide bonds than the target and therefore absorbs less per unit mass, so area normalization systematically under-reports it. Conversely, a scavenger adduct or a residual protecting group with strong absorbance can occupy more area than its mass share. These are second-order effects for well-made material, but they are the reason a chromatographic purity figure is characterized in the analytical literature as an estimate of relative composition rather than a mass fraction.

Integration parameters and the threshold that is rarely reported

After the separation and the detection comes the arithmetic, and the arithmetic has settings. Integration software applies a peak-detection threshold, a minimum peak area or height below which signal is treated as baseline noise and excluded from the summation. Raising that threshold removes small peaks from the denominator and raises the reported purity.

Baseline placement across a drifting gradient, decisions about whether to drop a perpendicular or apply a tangent skim to a shoulder peak, and whether the injection-front disturbance and any late-eluting wash peaks are included — each of these is an analyst-configurable choice, and each moves the final figure by a fraction of a percent. Between two competent laboratories, integration practice alone can account for several tenths of a percentage point.

This is a familiar situation in analytical chemistry rather than a defect peculiar to peptides. It does mean that differences of a few tenths of a percent between two purity figures carry little information, while a difference of several percent points to something more substantive — a genuinely different material, or a method with markedly different resolving power.

What HPLC purity does not establish

RP-HPLC separates by hydrophobicity, and a species that happens to share the retention behavior of the target will sit underneath the main peak regardless of how carefully the method is optimized. Several important classes of impurity are prone to this. Deamidation of asparagine or glutamine changes mass by approximately 1 Da and hydrophobicity very little. Aspartate isomerization changes mass not at all. An enantiomeric or epimeric impurity from racemization during synthesis is chromatographically similar on an achiral phase.

Detecting these requires an orthogonal method — mass spectrometry for the mass-shifted species, chiral or ion-exchange chromatography for the ones that are not mass-shifted. This is the reason a well-populated COA reports HPLC purity alongside a mass spectrometric identity confirmation rather than treating either as sufficient alone. The two techniques answer different questions: chromatography asks how much of the sample is the main component, and mass spectrometry asks whether the main component is the intended molecule.

Reading the number in context

A purity figure on a certificate of analysis is a summary statistic produced by a specific column, gradient, wavelength, and integration configuration. The figure is meaningful, and material characterized at 99% by a well-resolved method is meaningfully different from material characterized at 90%. But the precision implied by two decimal places exceeds the reproducibility of the measurement across laboratories, and small differences between vendors’ reported figures are better read as method variation than as material variation.

The more informative parts of an analytical package are often the ones that receive less attention than the headline percentage: whether the method is disclosed in enough detail to be reproduced, whether the chromatogram itself is included rather than only the derived number, whether an orthogonal identity method accompanies the purity determination, and whether peptide content is reported separately from chromatographic purity. Those elements describe the measurement rather than merely asserting its result, and they are what allow a reader to judge how much weight a given figure will bear.