qPCR Melt Curve Analysis: Peaks and Dimers

A melt curve (also called a dissociation curve) is the ramp your qPCR instrument runs after cycling finishes: it heats the plate slowly from roughly 60 °C to 95 °C while watching fluorescence fall as double-stranded product denatures. It exists for one reason — to tell you whether the signal you just quantified came from the product you intended to amplify, or from something else. This page is about reading it as a decision: what a given peak shape means, and what you should do next.

Before anything else, the limitation that decides whether this page applies to you at all.

Melt analysis only works with intercalating dyes — not hydrolysis probes

Melt-curve analysis requires a fluorophore that reports on any double-stranded DNA present in the tube. That is exactly what intercalating (DNA-binding) dyes such as SYBR Green I and the saturation dyes (EvaGreen, LCGreen, SYTO 9) do — and it is exactly why those chemistries need a melt curve in the first place: they cannot distinguish your amplicon from a primer dimer or an off-target product, because all three are double-stranded DNA.

Hydrolysis probe chemistry (TaqMan-style) does not produce a usable melt curve. The probe is cleaved during extension to separate reporter from quencher, so the accumulated fluorescence comes from free, already-separated reporter dye rather than from an intact duplex. Ramping the temperature afterwards has nothing to melt. This is not a defect: the probe is a third sequence-specific oligonucleotide that must hybridise for any signal at all, so specificity is built into the chemistry instead of being checked after the fact. If you are running a hydrolysis-probe assay and your instrument software offers a melt step, running it will not tell you anything about your product.

One real exception is worth knowing so you do not misapply the rule. Hybridisation probe formats — adjacent FRET probe pairs, and dual-labelled probes that remain intact — are compatible with melting analysis, but what melts is the probe-to-target duplex, not the amplicon. The peak Tm reports on how well the probe matched its binding site (which is how probe-based genotyping and mismatch detection works), and says nothing about whether an unrelated non-specific amplicon is also present. Do not read a probe-melt peak as an amplicon-specificity check; they answer different questions.

The practical rule: if your assay uses an intercalating dye, a melt curve is mandatory on every plate and every well, and it is the minimum evidence that your Cq values mean anything. See our broader guide to qPCR and RT-qPCR for how the dye-versus-probe choice affects the rest of the assay design.

What you are actually looking at: raw melt versus the derivative plot

Two plots come out of the same ramp, and they are read differently.

The raw melt curve plots fluorescence against temperature. It falls from a high plateau to near-baseline. A single transition appears as one sigmoidal drop; two products appear as two drops, which are hard to see by eye when they are close together.
The derivative plot plots the negative first derivative of fluorescence with respect to temperature (−dF/dT) against temperature. Each melting transition becomes a peak, and the temperature at the peak maximum is the melting temperature (Tm) of that transition. This is the plot everyone means when they say “melt curve”.

Peak height in the derivative plot is proportional to how much double-stranded DNA melted over that temperature interval — so it reflects both the amount of that species and how sharply it melts. It is not a quantitative measure of copy number, and comparing peak heights between wells is not a substitute for comparing Cq values.

Peak width matters more than most people use it for. A homogeneous amplicon of a single length and sequence melts over a narrow interval and gives a narrow peak. A broad peak, or a peak with a shoulder, means a population of duplexes melting over a range — either a mixture of species, or an amplicon with internal domains of very different GC content that melt in stages.

There is no universal “correct” Tm — it is a property of your assay

This is the single most common misreading. Tm is not a fixed number you can look up and check against. It is determined by:

GC content. G–C pairs carry three hydrogen bonds to A–T’s two, plus stronger base stacking, so a GC-rich amplicon melts higher. This is why very GC-rich targets sometimes need additives such as DMSO or betaine to denature reliably at all.
Amplicon length. Longer duplexes are more thermodynamically stable and melt higher, with diminishing returns as length increases. A 70 bp amplicon and a 200 bp amplicon of identical GC content will not share a Tm.
Sequence arrangement, not just composition. Nearest-neighbour stacking means two amplicons with the same GC percentage can melt at measurably different temperatures.
Buffer chemistry. Monovalent and divalent cation concentration shifts Tm — so master mixes from different suppliers, or a different Mg²⁺ concentration, will move it.
The dye itself. Saturation dyes bind more of the duplex than SYBR Green I at working concentrations and can shift and sharpen transitions differently.

The consequence for practice: the reference Tm for an assay is something you establish once, during validation, on a product you have independently confirmed — then treat as that assay’s expected value. Record it, along with the master mix and instrument used, in the assay’s documentation. A shift of roughly half a degree or more from that recorded value on a later run is a signal worth investigating; it is not evidence of anything on its own if you never established the baseline. In-silico Tm predictions from primer-design software are useful for expecting a rough range, but they are calculated under assumed conditions and should never be used as the acceptance criterion for a real run.

Reading the peak: what the shape means and what to do next

Work through the derivative plot for the sample wells and the no-template control together, not separately. The pattern below covers essentially everything you will see.

What you see	Most likely cause	What to do next
One narrow peak at the assay’s established Tm; flat NTC	Specific product, as expected	Proceed — but see the caveat below; one peak is consistent with specificity, not proof of it
A second peak roughly 5–20 °C below the product peak, usually broader and often also present in the NTC	Primer dimer	Confirm on a gel (dimers run as a diffuse band typically under ~100 bp). Raise annealing temperature, reduce primer concentration, set up on ice or use a hot-start polymerase, and redesign primers if it persists
A second peak close to or above the product peak	Non-specific amplicon of comparable or greater length/GC — mispriming, not dimer	Gel to size it, then sequence if the identity matters. Raise annealing temperature or redesign primers; check the primers against the actual genome, not just the target transcript
A broad single peak, or a peak with a shoulder that does not resolve	Either a mixture of products with similar Tm, or a genuinely heterogeneous amplicon (GC-domain structure, or an intended amplicon spanning a polymorphic site)	Do not resolve this from the melt curve. Run the product on a gel; a single clean band with a broad peak points to a real domain structure, multiple bands to a mixture
Correct product peak present in the sample, and also in the NTC	Template contamination of reagents or workspace — not dimer	Discard and remake the master mix, use filter tips, physically separate pre- and post-PCR work. Do not report the plate
Peak present in a no-reverse-transcription control at the product Tm	Genomic DNA carry-through in the RNA prep	DNase-treat, or redesign primers to span an exon–exon junction so genomic template cannot amplify
Peak Tm consistently shifted across a whole plate or a whole run	A changed condition, not a changed product: new master mix lot, different salt, different dye, or instrument calibration drift	Check what changed against the recorded validation conditions before assuming the amplicon is wrong. Re-verify against a known-good positive control
No peak at all, in wells that amplified	Melt ramp did not cover the Tm, or an analysis-software baseline/threshold problem	Check that the ramp end temperature is above the expected Tm and re-analyse; do not conclude anything about specificity from a missing curve
Low, ragged peaks only in high-Cq wells	Near the detection limit, where stochastic amplification and late artefact formation dominate	Treat as below the assay’s reliable range. Establish and report a limit of detection rather than reading these wells as results

Primer dimers: how they behave, and why a low peak is not automatically a dimer

Primer dimers form when primers anneal to each other — usually through complementarity at their 3′ ends — and are extended into a short, largely primer-derived duplex. Because they are short, they usually melt well below a real amplicon, and because they are a heterogeneous population of related products, their peak is often broad rather than sharp. They form preferentially where there is little or no target competing for primers, which is why the diagnostic move is always to look at the no-template control: a low-temperature peak present in the NTC as well as the sample is a dimer with near-certainty. A low-temperature peak in the sample only is not.

The temptation is to run the inference backwards — low peak, therefore dimer — and that is wrong often enough to matter. A short non-specific amplicon, an AT-rich off-target product, or a partially degraded product will also melt low. Conversely, non-specific products are not confined to low temperatures. Ruiz-Villalba and colleagues, characterising non-specific amplification systematically, reported artefacts in both classes for the same assay system: a low-Tm class around 78.5 °C against a specific product at 85.5 °C, and a separate high-Tm artefact class near 84 °C — close enough to the specific product to be easy to miss (Biomolecular Detection and Quantification 2017;14:7–18). Their broader finding is the one to carry: non-specific amplification occurred frequently and was unrelated to Cq or to calculated PCR efficiency, so neither of those numbers rescues you if the melt curve is ambiguous.

The correct statement is therefore narrower than the folklore: dimers melt low; a low peak is a reason to suspect dimer and check the NTC and a gel, not a diagnosis on its own.

Why one sharp peak does not prove specificity

A single narrow peak at the expected Tm is the result you want, and for routine run-to-run monitoring of a validated assay it is adequate evidence. It is not proof, for a structural reason: two different PCR products can share a melting temperature. Tm is a scalar summary of a duplex’s stability, and different sequences of similar length and GC content can land on the same value and co-melt as one apparently clean peak. MIQE 2.0 states the position plainly — melting analysis is more sensitive than size fractionation on agarose gels, but two different products can have the same Tm — which is why melt-curve evidence is treated as necessary, not sufficient.

Two products that differ in length but happen to match in Tm will resolve immediately on a gel, because gels separate on size where melt separates on thermal stability. That is the whole reason the two checks are complementary rather than redundant, and why the confirmatory step below is not optional at validation.

The confirmatory check: gel, sequencing, or both

Melt curve analysis is a per-run monitoring tool. Establishing that an assay amplifies what you claim it amplifies is a one-time validation exercise, and it needs orthogonal evidence.

Agarose gel electrophoresis answers “how many products, and what size?”. Run the post-qPCR reaction alongside a size ladder; you should see one band at the predicted amplicon size and nothing else. Dimers appear as a diffuse low-molecular-weight band, typically under about 100 bp, and are easy to distinguish from a real amplicon of a few hundred base pairs. A gel is cheap, fast, and the right first confirmatory step for any ambiguous melt curve. See agarose gel electrophoresis protocol basics for the mechanics, and note that a high-percentage gel (3–4%) or a precast small-fragment gel resolves short products far better than a standard 1% gel.
Sanger sequencing answers “is this the intended sequence?” — the only check that closes the same-Tm, same-size loophole. Gel-purify or column-purify the amplicon and sequence it in both directions. This is what you do once, when the assay is first validated, and again if you change primers, and it is what MIQE-conformant specificity evidence ultimately rests on. Our Sanger sequencing guide covers reading the trace and interpreting a mixed signal, which is itself informative: a superimposed double trace usually means you sequenced more than one product.
Ruiz-Villalba et al.’s recommendation is the same, and worth quoting as the working standard: for dye-based assays, melting curve verification is the minimal requirement, supplemented by gel electrophoresis and/or sequencing.

A defensible sequence for a new dye-based assay: design primers and check them against the genome rather than only the transcript; run a temperature gradient to find the annealing temperature that gives a single melt peak; confirm one band of the right size on a gel; sequence the product; record the Tm, master mix, and instrument as that assay’s reference; then rely on the melt curve for per-run monitoring against that recorded value.

Reading the controls, not just the samples

The melt curve of a control well carries more diagnostic information than the melt curve of a sample well, because you know what should be in it.

No-template control (NTC). Should be flat. A low-Tm peak means dimer; a peak at the product Tm means contamination. These have completely different remedies, and the melt curve is what separates them — an NTC that merely shows late amplification, with no melt data, cannot be triaged.
No-RT control. Should be flat in an RT-qPCR assay. A peak at the product Tm means genomic DNA is being amplified, which silently inflates every expression measurement on the plate.
Positive control. Its peak is the run’s internal reference for Tm. If the positive control has shifted along with the samples, suspect the run conditions; if only the samples have shifted, suspect the samples.

What MIQE expects you to report

The MIQE guidelines (Bustin et al., Clinical Chemistry 2009;55(4):611–622, doi:10.1373/clinchem.2008.112797) established the reporting framework: qPCR results are not interpretable by a reader unless the assay’s specificity evidence is disclosed alongside them. Melt-curve evidence is part of that disclosure for any dye-based assay.

MIQE 2.0 (Bustin et al., Clinical Chemistry 2025;71(6):634–651, doi:10.1093/clinchem/hvaf043, published 24 April 2025) is the current revision and changes the structure of the checklist itself: the original essential/desirable tiering is replaced by a single unified checklist answered Yes/No across five sections — reagent preparation, sample preparation, reverse transcription, qPCR protocol, and data analysis — with footnotes marking items that are highly desirable or only conditionally applicable. If you are citing MIQE in a methods section written now, cite the 2.0 revision, and be aware that a manuscript template built around the 2009 E/D designations is out of date.

In practice, for a dye-based assay, the specificity evidence a reviewer or auditor is entitled to see includes: the primer sequences and the amplicon length; the exact master mix and instrument (because both move Tm); the melt curve or the observed Tm with the ramp conditions used; the NTC result; and the orthogonal confirmation — gel image, sequencing result, or both — that established the product identity at validation. Depositing the melt-curve data alongside the amplification data, rather than only a Tm number in a table, is what makes the claim checkable; a data availability statement that covers raw qPCR output is the mechanism for that.

High-resolution melting is a different technique

High-resolution melting (HRM) uses the same physical event but is not the same assay, and conflating them causes real errors. HRM is a genotyping and variant-scanning method: it resolves single-base differences between amplicons by comparing the shape of normalised melting curves, not by reading a peak Tm off a derivative plot. It requires a saturating dye (SYBR Green I at working concentration is generally unsuitable because it redistributes during melting), a much finer temperature ramp, an instrument with the thermal precision and optical sampling to support it, and dedicated normalisation and clustering software.

Standard melt-curve analysis for specificity, which is what this page is about, needs none of that. If someone tells you your melt curve can distinguish two amplicons differing by one base, they are describing HRM on an HRM-capable instrument, not the post-run dissociation step on a general-purpose qPCR machine.

qPCR and RT-qPCR: how they differ, and how to run one properly — the parent assay guide, including reference genes, efficiency and controls.
PCR protocol basics — reaction setup, where dimer problems originate.
Agarose gel electrophoresis protocol basics — the size-based confirmatory check.
Sanger sequencing — the sequence-based confirmatory check.
RNA extraction protocol basics — where genomic DNA carry-through starts.
Thermal cycler calibration — why block temperature accuracy shows up as Tm drift.
Laboratory equipment and instrumentation — the wider cluster this page belongs to.

Frequently asked questions

What does a melt curve tell you in qPCR?

It tells you how many distinct double-stranded products were in the well at the end of cycling, and roughly how thermally stable each one is. It does not tell you how much of each there was, and it does not identify them by sequence — it separates species by melting temperature only.

What Tm should my qPCR product have?

Whatever your assay produced when you validated it. Tm depends on amplicon length, GC content, sequence arrangement, buffer salt concentration and the dye, so there is no target value that applies across assays. Establish it once on a confirmed product, record the conditions, and compare later runs against that.

How do I tell a primer dimer from a real product on a melt curve?

Look at the no-template control first. A low-temperature peak that appears in the NTC as well as the sample is a dimer. A low-temperature peak in the sample only could be a dimer, but could equally be a short or AT-rich non-specific amplicon — run a gel to size it before deciding.

Can I run a melt curve with a TaqMan probe assay?

No — hydrolysis probes are cleaved during extension, so there is no intact reporter-bearing duplex left to melt. Specificity in that chemistry comes from the probe having to hybridise for any signal to be generated. Hybridisation probe formats are a separate case: they do give melting data, but it reports on probe-target match, not on amplicon identity.

Does a single melt peak mean my qPCR is specific?

It means no product with a distinguishable Tm was present. Two different products can share a Tm and co-melt as one peak, so a single peak is necessary but not sufficient. Confirm identity by gel and sequencing at least once when the assay is validated.

Why does my melt peak have a shoulder?

Either two products are melting at close temperatures, or one amplicon contains domains of substantially different GC content that melt in stages. The melt curve cannot distinguish these; a gel can, because the two cases differ in the number of bands.

Why did my Tm shift compared to last month’s run?

Check what changed in the reaction before assuming the product changed. A new master mix lot, a different supplier, altered Mg²⁺, a different dye, or block calibration drift will all move Tm without any change in the amplicon. Re-run a known-good positive control to separate the two possibilities.

Should I report melt curves in a publication?

Yes, for any dye-based assay. MIQE 2.0’s checklist covers the qPCR protocol and data analysis disclosures that make specificity claims checkable, and depositing the melt data rather than only quoting a Tm is what lets a reader verify the claim rather than take it on trust.

A note on sources

The reporting requirements on this page are taken from the MIQE guidelines (2009) and the MIQE 2.0 revision (2025), cited above with DOIs. The quantitative observations about non-specific amplification are from Ruiz-Villalba et al., Biomolecular Detection and Quantification 2017;14:7–18, and are specific to the assays characterised in that study — treat them as illustrative of the two artefact classes, not as thresholds to apply to your own assay. Melt ramp rates, dye recommendations and HRM capability are instrument- and reagent-specific: consult your own instrument’s and master mix supplier’s technical documentation for the values that apply to your platform, rather than transferring numbers from another lab’s protocol.