The Cq is a single number, and a single number cannot tell you why it is what it is. The curve can. Nearly every common qPCR failure has a shape, and once you have seen a few of them you can diagnose a plate at a glance, long before you get to the analysis. This is a field guide to the curves that should make you stop and look: what each one is telling you about the reaction, and what to do about it. The reference point throughout is the healthy sigmoid covered in the companion piece on reading a curve; everything here is a departure from it.
A gallery of bad shapes
Most problems announce themselves as one of a handful of silhouettes. It is worth learning to recognise them as shapes first and reason about causes second.
**No amplification.** A flat line where you expected a curve. In a sample well this means the target was not there, was destroyed, or was never able to amplify: degraded template, a failed reverse-transcription step upstream, a pipetting miss that left the well without primers or polymerase, or an inhibitor carried in with the sample. Before you blame the biology, confirm the reaction was actually assembled. If a positive control on the same plate also failed, the problem is the assay or the setup, not the sample.
**Late, weak rise.** A curve that lifts off far later than expected and never really commits to a steep climb. This is the signature of very little starting template, but also of a partially inhibited reaction or a suboptimal annealing temperature. The way to tell them apart is a dilution: if diluting the sample makes the Cq earlier, or at least does not delay it as much as dilution should, an inhibitor was holding the reaction back and you have just diluted it out.
**Jagged or noisy.** A curve that climbs but shivers as it goes, or a trace that is spiky throughout. This usually points at the detection side rather than the amplification: too little probe, an unstable passive reference dye, air bubbles in the well, or inconsistent mixing. It can also be a genuinely weak signal amplified by the software into visible noise. Check the reference dye and make sure wells are properly sealed and mixed before you touch the chemistry.
**Low plateau.** A curve of normal timing that tops out well below its neighbours. The reaction started fine but ran out of something early: limiting or degraded reagents, or a probe whose signal is quenched. A low plateau on its own rarely corrupts the Cq, since the Cq is read in the exponential phase, but a plateau that is low across the whole plate is a hint that the master mix or an instrument setting is off.
**Amplification in the no-template control.** The most important curve on the plate is the one that should stay flat. An NTC that amplifies means contamination: stray template in a reagent, aerosol from a previous run, or amplicon carried over on the deck. The fix is procedural, not analytical. Bleach the workspace, open fresh reagents, separate reaction setup from post-amplification handling, and if the NTC climbs very late with a low melt temperature, suspect primer-dimer rather than true contamination.
**The hook.** A curve that rises normally and then bends back down at the top instead of holding a plateau. The hook effect comes from an enormous amount of starting template overwhelming the reaction and the detection chemistry, and it is a cousin of the inverted curve discussed below. The remedy is to dilute the sample and run it again.
The inverted curve, and why it fools you
One shape deserves its own diagram because it is genuinely counterintuitive: a curve that starts by sloping downward, dips below zero, and only then shoots up. The instinct is to read this as a broken well. It is usually the opposite: a well with far too much template.
The dip is an artefact of baseline subtraction. When a sample starts with an extremely high copy number, real amplification begins in the very first cycles, inside the window the software has chosen to treat as flat baseline. The baseline fit is pulled up by that early signal, and when it is subtracted from the curve, the early cycles are pushed below zero. The result is a curve that appears to lose signal before it gains it, followed by a crossing that lands far earlier than it should. The danger is that the Cq still looks like a number, just a very low one, so it slips through unless you look at the shape.
Two fixes apply. Dilute the sample, often heavily, a thousandfold or more, so amplification no longer starts inside the baseline window; the curve straightens out and the Cq becomes meaningful. Or, if the software allows it, correct the baseline window by hand so it ends before real signal appears. Diluting is the more reliable of the two, because it fixes the reaction rather than just its interpretation.
Reading shapes before you read numbers
The thread running through all of these is that the curve carries a diagnosis the Cq cannot. A number tells you where a crossing landed; the shape tells you whether the crossing meant anything, and if not, why. Get in the habit of scanning the amplification plot before you export a single value: look at the no-template controls first, then the overall family of curve shapes, then the individual outliers. Most of what you need to know is visible in ten seconds of looking.
A Cq can only ever say when a curve crossed the line. The shape of the curve is what tells you whether the crossing was real.
A useful discipline is to keep the healthy silhouette in mind as a template and treat every deviation as a question rather than a verdict. A late curve asks whether there was little template or an inhibitor. A jagged one asks about the dye and the seal. A climbing no-template control asks whether it is contamination or dimer. The plate rarely answers on its own, but the shape tells you which question to ask, and that is most of the battle.