What the 2^-ΔΔCt method actually measures
The 2−ΔΔCt method converts four cycle numbers into one statement: how much more, or less, of a transcript your treated sample contains compared with a control sample. It is a relative quantification. It never tells you how many copies are present — for that you need a standard curve of known copy number, which the DNA copy number calculator helps you build.
The logic runs in two subtractions. The first, ΔCt, subtracts a reference gene's Ct from the target gene's Ct within a single sample. That cancels everything that affected the whole sample equally: how much RNA you loaded, how well the reverse transcription went, pipetting volume, inhibitor carry-over. The second subtraction, ΔΔCt, takes the treated sample's ΔCt minus the control's. That cancels everything intrinsic to the two assays — primer efficiency, amplicon length, probe chemistry — because both assays appear on both sides.
What survives both subtractions is the biology. Raising 2 to the power of the negative of that number converts a cycle difference into a concentration ratio, because each PCR cycle that a template starts earlier means it was present at twice the amount. One cycle is a factor of two; ten cycles is a factor of 1,024.
The minus sign catches people out. A lower Ct means more template, so a negative ΔΔCt produces a fold change above 1, and a positive ΔΔCt produces a fold change below 1.
Where the formula comes from, variable by variable
During the exponential phase of PCR, the amount of product after n cycles is N0 × En, where E is the amplification factor — 2 for a perfect reaction that doubles every cycle. The instrument reports the cycle at which product crosses a fixed fluorescence threshold. Two wells that cross the same threshold contain the same amount of product at their respective Ct values, so the ratio of their starting amounts is E raised to the difference in their Ct values.
ΔCt normalises within a sample. Because Ct is a logarithm (base E) of the starting quantity, subtracting Ct values is the same as dividing quantities. ΔCt = Cttarget − Ctref is therefore the log-ratio of target to reference in that tube — a dimensionless quantity that no longer depends on how much cDNA you pipetted.
ΔΔCt compares samples. Subtracting the control's ΔCt from the treated ΔCt gives the log of the ratio-of-ratios. Exponentiating with a minus sign, 2−ΔΔCt, returns that ratio-of-ratios on a linear scale, with the control fixed at exactly 1 by construction.
E is the assumption you are making. Livak and Schmittgen's derivation requires that the target and reference assays amplify with the same, near-perfect efficiency. The Pfaffl formulation drops that requirement: it uses each assay's own measured amplification factor, obtained from the slope of a standard curve as E = 10−1/slope. A slope of −3.32 gives E = 2.00 exactly, which is where the familiar “slope of −3.3 means 100% efficiency” rule comes from. Fit that curve with the qPCR standard curve efficiency calculator before you decide which model to use.
Worked example: a target induced 4.9-fold
You treat cells for six hours, harvest RNA from treated and untreated flasks, and run your target gene and GAPDH in triplicate on both. The mean Ct values are the defaults loaded above:
- Target, treated: 22.50; GAPDH, treated: 18.00
- Target, control: 25.00; GAPDH, control: 18.20
- ΔCt for the treated sample. 22.50 − 18.00 = 4.50 cycles.
- ΔCt for the control sample. 25.00 − 18.20 = 6.80 cycles.
- ΔΔCt. 4.50 − 6.80 = −2.30 cycles. Negative, so the target is more abundant after treatment.
- Fold change. 2−(−2.30) = 22.30. Since 22 = 4 and 20.3 = 1.2311, the product is 4 × 1.2311 = 4.925×.
- log2 fold change. This is simply −ΔΔCt, so +2.30. That is the number you would plot on a volcano plot axis.
Read the result as: after treatment, the target transcript is present at 4.9 times its control level, once differences in RNA input have been removed by GAPDH normalisation.
Now check the reference gene. GAPDH moved from 18.20 to 18.00, a shift of 0.20 cycles, which corresponds to a factor of 20.20 = 1.149 in apparent GAPDH abundance. That is within the noise you should expect from pipetting and RT variation. Had GAPDH moved by 1.5 cycles instead, the same arithmetic would have quietly folded a factor of 21.5 = 2.83 of normaliser drift into your answer.
How to read the fold change you get
A fold change of 1.000 means no detectable difference. Above 1 the target is more abundant in the treated sample; below 1 it is less abundant. The scale is multiplicative and asymmetric on a linear axis: a doubling is 2.0 while a halving is 0.5, which is why expression data is nearly always reported and plotted as log2 fold change, where doubling is +1 and halving is −1 and the two are symmetric about zero.
How big is big enough to believe? The honest answer is that it depends on your replicate scatter, not on a universal cut-off. The arithmetic gives you a hard floor to reason from: the fold change is 2 raised to a cycle difference, so an uncertainty of ±0.5 cycles in a single Ct value multiplies or divides the answer by 20.5 = 1.41. An uncertainty of ±1 cycle is a factor of 2. The sensitivity table this calculator returns shows exactly that for your own numbers. If your technical replicates spread over half a cycle, a 1.4-fold result carries no information at all.
Two consequences follow. First, run enough biological replicates — independent flasks or animals — and do statistics on the ΔCt values, which are approximately normally distributed, rather than on fold changes, which are not. Second, report the fold change with a confidence interval or the range across replicates. The MIQE guidelines make this explicit: a relative quantification result without replicate structure, efficiency values and normaliser justification is not interpretable by anyone else.
Beware the direction convention too. “Down 4-fold” and “0.25-fold” describe the same result; publications use both. This calculator always reports the ratio treated ÷ control, so a repressed gene comes back as a number between 0 and 1.
ΔΔCt converted to fold change
| ΔΔCt (cycles) | Fold change (2^−ΔΔCt) | log₂ fold change | Plain reading |
|---|---|---|---|
| −5.00 | 32.000 | +5.00 | 32× up |
| −3.32 | 10.000 | +3.32 | 10× up (one log₁₀) |
| −2.00 | 4.000 | +2.00 | 4× up |
| −1.00 | 2.000 | +1.00 | doubled |
| −0.50 | 1.414 | +0.50 | 41% up |
| 0.00 | 1.000 | 0.00 | unchanged |
| +0.50 | 0.707 | −0.50 | 29% down |
| +1.00 | 0.500 | −1.00 | halved |
| +2.00 | 0.250 | −2.00 | 4× down |
| +3.32 | 0.100 | −3.32 | 10× down |
| +5.00 | 0.031 | −5.00 | 32× down |
3.32 cycles is log₂(10), which is why a ten-fold dilution series steps a standard curve by about 3.3 cycles when efficiency is 100%.
The MIQE guidelines and what they ask you to report
The Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines, published by Bustin and colleagues in Clinical Chemistry in 2009, set the reporting standard this calculator is built around. For a relative quantification result they ask for, at minimum: the amplification efficiency of every assay and how it was determined, the identity and validation of every reference gene, the number and type of replicates, the RNA quality assessment, and the exact quantification model used. A fold change quoted without those is not reproducible.
The practical consequence for this page: if you cannot state your efficiencies, you cannot honestly claim the Livak model applies, because the Livak model is the assumption that both efficiencies equal 100%.
Mistakes that produce a wrong fold change
- Swapping treated and control. The two ΔCt values enter the subtraction in a fixed order. Reversing them inverts the answer — a 4-fold induction becomes a 0.25 result.
- Averaging fold changes across replicates. Average the ΔCt values, then exponentiate once. Averaging the exponentiated values inflates the mean, because the arithmetic mean of a set of ratios exceeds their geometric mean whenever the ratios differ.
- Using an unvalidated reference gene. GAPDH and ACTB are both regulated under hypoxia, in the cell cycle, and in many differentiation protocols. Validate that yours does not move under your specific treatment.
- Applying Livak to assays with mismatched efficiency. If the target amplifies at 85% and the reference at 105%, the base of 2 is wrong for both, and the error compounds with the size of the Ct difference.
- Including wells that never left baseline. A Ct of 40 reported by the instrument for a well with no amplification curve is not a measurement; it is a censored value. Exclude it and say so.
- Comparing Ct values across plates without a calibrator. Threshold placement and baseline settings differ between runs. Carry the same calibrator sample on every plate.
- Quoting three decimal places on a ratio built from Ct values that scatter by half a cycle. The precision of the output is set by the scatter of the input, not by the calculator.
When to use a different method instead
Use the Pfaffl model whenever your standard curves show that the target and reference assays amplify at different rates. It is the same idea with each assay's own base instead of a shared 2, and it collapses back to the Livak answer exactly when both efficiencies are 100% — which is a useful check that you have entered the numbers correctly.
Use an absolute standard curve when you need copies rather than ratios: viral load, gene dosage, copy number variation, or any result that must be compared with a number produced in another laboratory. That means running a dilution series of a template of known concentration on the same plate. Prepare it with the serial dilution calculator and quantify the stock with the A260 concentration calculator.
Use digital PCR when the expected difference is smaller than about 1.5-fold, when inhibitors are unavoidable, or when you have no acceptable reference gene. Partitioning the reaction and counting positive droplets gives an absolute count that does not depend on amplification efficiency at all.
Use multiple reference genes — the geometric mean of two or three validated normalisers — for any experiment where a single housekeeping gene might respond to the treatment. The arithmetic is unchanged; you simply substitute the geometric mean Ct for the single reference Ct. Before designing new normaliser assays, check their melting temperatures match the rest of your panel so one annealing temperature suits every well.
Key terms
- Ct (Cq)
- Quantification cycle: the fractional cycle number at which the amplification curve crosses a set fluorescence threshold. MIQE prefers the term Cq; instruments still print Ct.
- Calibrator
- The sample the fold change is expressed relative to — here, the control. Its fold change is 1.000 by definition, not by measurement.
- Reference gene
- A transcript assumed to be unaffected by the treatment, used to cancel differences in RNA input and reverse transcription. Also called a housekeeping gene or normaliser.
- Amplification efficiency
- The fraction of template copied per cycle, reported as a percentage. 100% means a perfect doubling; it is derived from the slope of a standard curve as E = 10^(-1/slope) - 1.
- log₂ fold change
- The base-2 logarithm of the fold change, equal to −ΔΔCt under the Livak model. Symmetric about zero, which is why it is the standard axis for expression plots.