Guides & comparisons
Reading the Chromatogram: What a Purity Percentage Hides
If you buy research peptides, read the chromatogram, not just the number: an HPLC purity percentage summarises that picture, and peak shape, shoulders, the baseline and the integration windows decide what it means — a clean-looking 99% can still describe the vial poorly.
“99.1% by HPLC” is a summary of a picture. The picture is usually on page two of the certificate and almost nobody looks at it — which is a problem, because the number can be entirely accurate and still be a poor description of what is in the vial. You don’t need chemistry to read the picture; you need to know what the axes mean and where the judgement calls hide.
What does an HPLC trace actually plot?
The x-axis is retention time in minutes; the y-axis is detector response, usually in milli-absorbance units (mAU).
Analytical peptide HPLC is nearly always reversed-phase: a nonpolar stationary phase, commonly C18-bonded silica, and a mobile phase that starts water-rich and grows richer in an organic solvent — typically acetonitrile — with a little acid as modifier. More hydrophobic molecules hold onto the column harder and need more organic solvent to come off, so they elute later. Retention time is a proxy for hydrophobicity, and molecules that resemble each other chemically elute close together. Hold onto that last clause: it is the whole problem.
Where does the purity percentage come from?
Software draws a baseline beneath each peak, integrates the area between peak and baseline, sums the areas it recognises, and divides: main peak area ÷ total integrated area × 100. Three things there are decisions rather than measurements — which baseline, which peaks, and what counts as the total. Each is a legitimate place for the number to drift.
What does the peak shape tell you?
The ideal peak is symmetric and roughly Gaussian. Real peaks deviate, and the deviations read.
Tailing — a steep rise and a slow return dragging along the baseline — is common in peptide work, usually from secondary interactions with the stationary phase. Mild tailing is normal. Severe tailing matters for a purity figure specifically, because the tail occupies exactly the region where a slightly-more-retained impurity would appear, and it can absorb one entirely. Asymmetry is quantified as a tailing factor: near 1.0 is symmetric, and values well above roughly 1.5 deserve a second look.
Fronting — slow rise, sharp fall — usually means column overload, which is worse than it sounds for a purity assay. The main peak broadens into the region where small peaks live and the detector may be outside its linear range at the apex: the main peak gets fat, the impurities get relatively smaller, and the reported purity goes up. Broad, rounded peaks throughout suggest a tired column or a poor gradient, and weak resolution inflates purity the same way — by merging things that should have been separated.
Why do shoulders on a peak matter? The deletion-sequence problem
This is the most important section here.
Peptides are assembled stepwise, and the characteristic synthesis failure is a deletion sequence — a chain missing one residue because a coupling step didn’t go to completion. The result is nearly the same size and nearly the same hydrophobicity as the target, so it elutes very close to it.
“Very close” frequently means not resolved into its own peak. It shows up as a shoulder on the leading or trailing flank of the main peak, or as a subtle asymmetry, or as nothing visible at all while still being present. And because integration treats a shoulder as part of the main peak unless someone deliberately splits it, that material commonly gets counted as product.
So a 99% trace with a visible shoulder is weaker evidence than a 97% trace with four clean, well-separated small peaks: in the second case you know what you’re looking at, and in the first the number is carried by a peak that may hold more than one species. Look at both flanks of the main peak, deliberately, every time.
What should the baseline look like?
Look at the flat regions before the first peak and after the last.
Noise — the fuzz on the baseline — sets the detection limit. An impurity whose peak is smaller than the noise is not absent, it is invisible; a visibly noisy trace has a high floor for what it can claim isn’t there.
Drift — a baseline climbing through the run — is normal in gradient work, because the organic solvent and acid modifier absorb differently as their proportions change. Modest drift is expected. Heavy drift matters because the software must decide where the baseline sits under each peak, and on a steep slope that decision changes the areas.
How can integration windows inflate a purity figure?
None of this is forgery. All of it is legitimate practice that can flatter a number.
Run truncation. If the gradient runs to twenty minutes and the trace ends at twenty minutes, anything more hydrophobic than the final solvent composition never eluted and was never counted. A well-designed method holds at high organic at the end and shows that hold — a flat, empty stretch after the last peak. If the trace stops the moment the main peak comes down, you’re looking at a window, not a run.
The solvent front. Everything eluting in the void volume, the first minute or two, is unretained: salts, counterions, small polar material. Excluding it is standard and reasonable. It also means “100% of the integrated area” excludes a whole category of vial contents.
Manual reintegration. Peaks can be excluded by hand and baselines redrawn — not inherently wrong, but invisible in a printed image. What makes it visible is a peak table: retention time, area and area percent for every peak found. If the table sums to 100 and its rows match what you can see, the integration was at least disclosed. A chromatogram image with no peak table is a picture, not data.
Why does a clean-looking trace prove less than it seems?
The y-axis is almost always scaled so the main peak fills the frame. If that peak is 2000 mAU tall, a 1% impurity is 20 mAU — one hundredth of the plot height, a barely visible bump — and a 0.2% impurity is indistinguishable from the axis line.
That is why an apparently empty baseline is weak evidence of purity. The honest presentations are a zoomed inset of the baseline region, a peak table, or both. A full-scale image plus a number is unfalsifiable by construction: it cannot be checked, so it doesn’t function as evidence.
Why does the detection wavelength matter?
214 nm targets the peptide bond, so essentially every peptidic species responds, roughly in proportion to how many bonds it has. 220 nm is close and commonly used. Both are defensible.
280 nm is a different setting entirely: it detects aromatic side chains — tryptophan, tyrosine, weakly phenylalanine. A peptide containing none of those barely absorbs there, and impurities lacking them are invisible regardless of quantity. A high purity figure at 280 nm on a sequence with no aromatic residues is close to meaningless — a reason to read sequence and wavelength together.
And at any UV wavelength, species that don’t absorb — inorganic salts, many residual solvents — never appear at all. They aren’t impurities the chromatogram missed; they’re impurities the technique cannot see. Hence residual solvents getting their own test, usually by gas chromatography, and hence several relevant questions being absent from a purity COA.
Quick reference: what to check on a chromatogram
| What you see | What it suggests |
|---|---|
| Shoulder on the main peak | Co-eluting related species counted as product |
| Heavy tailing | The region where a late impurity would hide is obscured |
| Fronting or a flat-topped peak | Column overload; purity likely overstated |
| Noisy baseline | High detection floor — “nothing there” means “nothing above the noise” |
| Trace ends right after the main peak | Late-eluting material may never have been counted |
| No peak table | Integration undisclosed; the number can’t be checked |
| No axis labels or retention times | An image, not a chromatogram |
| Identical traces across two lots | One analysis, two certificates |
That last row is a different category of problem, and it belongs with fake, recycled and mismatched certificates — as does the prior question of whether the laboratory whose name sits above the trace is real and accredited for the method.
How much is an HPLC purity figure worth?
Reading a trace won’t make you an analyst, and it won’t turn an unverifiable document into a verifiable one. What it does is tell you when a number is doing more work than the data behind it can support — which is most of the useful skepticism available to a non-chemist.
The follow-through is unglamorous: keep the trace, not just the figure, attached to the lot it describes, so a question six months from now has something to look at. That’s why Peppies attaches certificates to individual shelf items rather than to product names. See also how to read the rest of the certificate and why verification belongs in inventory.
A chromatogram describes one injection of one sample on one day. Nothing more, and we won’t claim more for it. Research use only.