What “98% by HPLC” Is a Percentage Of
Picture a vial whose certificate reads 99 percent. Weigh ten milligrams of the powder out of it and you may well be handling seven or eight milligrams of actual peptide. The certificate has not lied. It answered a narrower question than the one most readers arrive with, and that gap is where the trouble starts.
Four operations on a trace
Deriving the figure takes almost no arithmetic. Integrate each peak. Sum those areas. Divide the largest by the sum. Multiply by a hundred. Weight enters nowhere in that sequence.
A detector registers molecules crossing its flow cell, scaled by how strongly each absorbs at whatever wavelength was selected. The quotient therefore sets one detector response against the sum of all of them. Converting it into a fraction by weight would require every species present to absorb the same amount per unit of mass. None of them do.
Everything the denominator cannot see
To join the total, a species has to travel through the column and then absorb light on its way past the detector. Whatever fails either condition is simply absent, and several things fail routinely.
Residual moisture is one. Freeze-dried peptide commonly retains several percent water, which has no chromophore whatsoever. Counter-ion is another: trifluoroacetate or acetate, which can account for a meaningful share of whatever sits on the balance, yet never reaches the chromatogram. Then come buffer components, leftover reagents and other salts that carry nothing for the detector to register. Last, anything that never leaves the column at all counts for nothing, whether it stuck irreversibly to the stationary phase or aggregated into particles too large to travel.
Hence the vial in the opening paragraph. Ninety-nine percent by area is entirely compatible with a good deal less than ninety-nine percent peptide by weight, because separate techniques answer the separate questions. Mass fraction comes from amino acid analysis or nitrogen determination. Water comes from Karl Fischer titration. Area percentage comes from HPLC. Reporting any one of the three says nothing about the other two.
Same vial, two laboratories, two answers
Split a batch, send half to each of two competent labs, and the returned percentages can differ with neither analyst having erred. They computed different denominators.
Flatten the gradient and species that previously merged now separate, so the total gains peaks. Shift the detection wavelength and the roster of contributors changes, along with how much each contributes. Extend the run and late material appears that a shorter method never waited for. Swap column chemistry and a different subset of impurities resolves. Every one of those choices alters the sum sitting underneath the main peak, and therefore alters the percentage printed on the page.
Someone decides where the baseline goes
Integration involves judgement, not only computation. An analyst chooses where to run the baseline beneath a peak. An analyst decides whether a shoulder becomes its own peak or gets folded into the one beside it. Both choices move the result.
Worse, the influence peaks precisely where accuracy matters: on main peaks that come out broad or shouldered, which is what proline-rich sequences tend to give. Several tenths of a percent can separate two experienced analysts working the same trace, and that spread is the realistic precision limit on the whole figure. More on this in how integration choices move a purity number.
When one peak is really two compounds
Built into the calculation is a premise that each peak corresponds to a single species. Break that premise and the consequences run in one direction only. Anything eluting at the same moment as the target is tallied as target, so the reported purity goes up rather than down.
That asymmetry makes co-elution the failure worth worrying about most. It flatters the material, and it leaves the chromatogram looking perfectly ordinary. Staring harder at the same trace will not reveal it; only a second method working on an orthogonal principle will. The mechanism gets a full treatment in what co-elution does to a purity figure.
Why 98 percent here and 98 percent there differ
Response is not uniform across species, and the direction of the resulting distortion depends on the wavelength in use and on which impurities happen to be present.
| Wavelength | Response tracks | What gets understated |
| 214 nm | Count of backbone amide bonds | Short truncated species, which deliver less area per mole than intact material and are systematically under-represented |
| 280 nm | Aromatic residue content | An impurity that has lost the sole tryptophan, which can be present in full while registering almost nothing |
Neither choice yields a proportional portrait of the mixture. The argument is laid out in why two peptides at 98 percent are not equally pure.
The job the number does well
None of the above condemns the measurement. Within a fixed method it is sensitive and reproducible, it costs little enough to run against every lot, and it catches the impurity families that solid-phase synthesis genuinely throws off: deletions, truncations, oxidations, deprotections that did not go to completion.
Its legitimate use is narrow and useful. Compare one lot of a compound against another lot of that same compound, analyzed the same way. Inside that comparison the figure behaves exactly as designed, and it is trustworthy.
What has to accompany the figure
Wavelength. Gradient. Column. Run length. Given those four, a reader can reconstruct what went into the denominator. Missing them, the number is a share of an unstated total, and setting it beside a number from some other laboratory sets two unrelated calculations side by side. Further reading: reading an HPLC chromatogram and net peptide content explained.
