Reading a chromatogram: peaks, baseline, signal-to-noise and calibration curves

A chromatogram is the graph an HPLC instrument draws as a sample passes through it. The purity figure on a certificate of analysis (COA) is worked out from that graph. Being able to read it lets you check the number instead of taking it on trust. This guide goes through a chromatogram piece by piece: the axes, the peaks, how they are measured, how small a peak can be and still count, and how labs turn peak sizes into amounts.

For how the instrument separates compounds, see how HPLC works. For what a purity percentage does and doesn't tell you, see HPLC purity explained. Terms are defined in the glossary.

What is a chromatogram?

A chromatogram is a plot of a detector's signal against time, with one peak for each component that comes off the column. Reading one starts with its two axes:

• The horizontal axis is time, usually in minutes from the moment the sample was injected.
• The vertical axis is detector response. With a UV detector this is absorbance, usually in milli-absorbance units (mAU). Peptides are typically measured at 210–230 nm, where the peptide bond absorbs.

The flat line where nothing is coming off the column is the baseline. Every peak is measured from it.

The parts of a peak

Each peak is one component, or several that weren't separated, leaving the column.

The solvent front

Most chromatograms start with a disturbance in the first minute or two, when the injection solvent and anything that doesn't stick to the column pass through. This is the solvent front, or void. It isn't counted as part of the sample. A component that elutes inside it can't be measured, which is one reason a purity figure may miss very polar impurities.

Peak area and integration

The software measures peak areas by integration: it decides where each peak starts and ends, draws a baseline under it and adds up the area. The result is in units such as mAU·s.

Integration involves choices, and they change the result:

• Where the baseline is drawn. If the baseline drifts during a gradient, a line drawn too high cuts off area; too low adds it.
• How overlapping peaks are split. A small peak on the side of a large one can be split off with a vertical drop line, or skimmed off with a tangent. The two give different areas.
• The threshold. Peaks below a set size may be ignored entirely.

A good COA shows the trace with the integration marks on it, so you can see what was counted.

Working out area-percent purity

Here is an illustrative peak table, of the kind printed under a chromatogram:

The purity is the main peak's area divided by the total: 1,985.6 ÷ 2,012.0 × 100 = 98.69%.

This method assumes that every component gives the same signal for the same amount. At low UV wavelengths that is roughly true for peptides, because most of the absorbance comes from peptide bonds. It is less true when an impurity has extra light-absorbing groups, such as a tryptophan residue the main peptide lacks. For what the small peaks usually are, see reading impurity peaks.

Resolution: are the peaks separated?

Resolution (Rs) measures how well two neighbouring peaks are pulled apart. It compares the gap between their retention times with their widths:

Rs = 2 × (t₂ − t₁) ÷ (w₁ + w₂)

where t is each peak's retention time and w its width at the base.

• Rs of 1.5 or more means the two peaks are separated down to the baseline.
• Rs around 1 means they overlap, and their areas can't be measured cleanly.

If an impurity isn't resolved from the main peak, it is counted as part of it, and the purity reads high. A shoulder on one side of the main peak is the visual clue.

Peak shape

An ideal peak is symmetrical. Two common distortions:

• Tailing: the back of the peak drags out. It often comes from the peptide interacting with the column in more than one way, or from an ageing column.
• Fronting: the front of the peak slopes. It often means too much sample was loaded.

Labs measure symmetry with a tailing or symmetry factor, where 1.0 is perfect. Methods usually set an acceptable range as part of their checks before running samples.

How do you calculate signal-to-noise ratio?

Even with no sample, the baseline isn't perfectly flat. It wobbles slightly because of electronic noise, pump pulses and small temperature changes. Signal-to-noise ratio (S/N) compares a peak's height with that wobble.

The European and US pharmacopoeias calculate it as:

S/N = 2H ÷ h

where H is the peak's height above the baseline and h is the range of the baseline noise over a stretch with no peaks (European Pharmacopoeia, chapter 2.2.46).

S/N decides whether a small peak is real:

These are the usual working values, and the ICH guideline on validating analytical methods describes the signal-to-noise approach (ICH Q2(R2)). An impurity below the LOQ might be listed as "less than 0.05%" rather than given a number, or not listed at all. So a trace that looks clean is clean only down to the method's detection limit.

What is a calibration curve?

A calibration curve is a graph of a method's response, such as peak area, against known concentrations of a reference standard, used to read off the concentration of an unknown sample. Area-percent purity gives the share of the peaks. It doesn't say how much compound is in the vial. For that, a lab needs a calibration curve, also called a standard curve.

1. Prepare reference standard solutions at several known concentrations. A reference standard is a well-characterised sample of the same compound.
2. Run each one and measure the peak area.
3. Plot area against concentration. Over the method's working range the points should fall on a straight line.
4. Draw the line of best fit. Software fits it by least squares: it picks the line that makes the squared vertical gaps between the points and the line as small as possible (NIST/SEMATECH e-Handbook).
5. Run the sample and read its concentration off the line.

Things to check on a calibration curve:

• Linearity. The fit is usually reported as a correlation coefficient (r) or r², which should be very close to 1. A value of 0.999 or better is common for HPLC.
• Range. The sample's result must fall between the lowest and highest standards. Reading beyond them assumes the line stays straight, and it may not.
• The standard itself. The answer is only as good as the reference standard's own stated purity and content.

This is how assay and content figures are measured. It is one way of working out net peptide content, which purity alone can't give. For the same idea applied to UV absorbance, see measuring peptide concentration and the Beer–Lambert calculator.

Relative standard deviation and system checks

Before running samples, labs inject a standard several times to confirm the instrument is behaving. This is a system suitability check, one of the routine controls a lab's quality system sets; see QA vs QC. A typical requirement is that the peak areas from replicate injections agree within a stated relative standard deviation (RSD), often 2% or less, along with limits on resolution and tailing.

RSD is the standard deviation of the results as a percentage of their mean. A low RSD means the instrument gives the same answer each time. For how to calculate it, and how to judge whether two results really differ, see accuracy, precision and significant figures. For putting a range around the mean of repeat results, see confidence intervals and standard error.

Blank runs and ghost peaks

A blank is a run with no sample, only the solvent. Any peak in the blank comes from the system, the solvent or the water, not the sample. Comparing the blank with the sample run shows which peaks to ignore. Poor-quality water is a common cause of these "ghost peaks"; see lab water.

A checklist for a COA chromatogram

When a COA includes the trace, check:

• The axes and conditions: wavelength, column, gradient and run time are stated.
• The main peak: one dominant peak, with no obvious shoulder.
• The baseline: flat either side of the peaks, and the integration marks look sensible.
• The peak table: areas add up, and the main peak's area % matches the stated purity.
• Small peaks: listed with their own area %, or covered by a stated detection limit.
• A blank, if shown: peaks in it aren't counted in the sample.

For a full walk-through of a lot document, see how to check a certificate of analysis. Every lot has its own COA, published on the product page and on that lot's own page. You can download the COA for any lot from the COA index and read it against this checklist. See Quality for how we document our products.