Skip to content

Use it for · Gene expression

Fold changes from your Ct values, with their error and a test

Bring last week’s run file. Clarida, qPCR analysis software, turns its Cq values into fold changes against your control group, normalized on your reference genes.

4 in your first month, then 2 a month. No card.

New experimentqPCR start page › Relative quantification
Already ran the plateAnalysis only › Import & review
On this page

What you get

Every target in each sample gets a normalized relative quantity with its standard error. Scale it to your control group and it is the fold change you report. The error is carried from the replicate Cq values, through each target’s efficiency and the reference genes, to the normalized quantity.

Beside it you see the efficiency each target used and where it came from, and how stable your reference genes were, as their geNorm M and coefficient of variation. Test your groups and each target gets its fold change with a 95% confidence interval. The numbers leave in your Excel notebook, and the saved analysis goes into a PDF or HTML file with the Report builder.

no change0.5124816Fold change, LPS vs control (log scale)IL66.5× 95% CI 4.8 to 8.9IL1BTNFCCL2no change124816Fold change vs control (log)IL66.5× 95% CI 4.8 to 8.9IL1BTNFCCL2
Fig. 1Fold change after LPS for four targets, each with its 95% confidence interval, n = 3 per group: none of the intervals touches no change. Illustrative data.

Is it your ΔΔCt?

The classic 2−ΔΔCt is the special case of what Clarida computes. Each row below adds accuracy; none changes what the number means.

Efficiency 2 for every target, one reference gene, scaled to the control groupThe fold change is your 2−ΔΔCt
Each target’s own efficiencyFrom its standard curve, your library or a value you enter; 2 only when none is known, and Clarida says so
Several reference genesNormalized on their geometric mean, with their stability shown beside them

How it works in Clarida

01In the app: Design › Assay Prep

Mark which genes are references

Normalization reads each target’s type, so the reference genes you tested for stability are used without being picked again. A target taken from your library keeps the type you gave it there.

  • Default type: gene of interest; reference gene is one of five types

In full on the Design page

02In the app: Analyze › Cq processing › Efficiency

Each target corrected for its own primer efficiency

A target that amplifies 1.9-fold per cycle is off by about 40% over a ΔCq of 7 if you assume perfect doubling. So each target takes its efficiency from a standard curve in this experiment, whose error then flows into every quantity, or from your library.

  • Default: automatic, this experiment’s standard curve first, then your library, then 2
  • Per target: override any of these

In full on the Analyze page

03In the app: Analyze › Cq processing › Normalization

Normalized on your reference genes

Each sample’s normalization factor is the geometric mean of its reference genes’ relative quantities. A sample missing one of them gets no normalized value and is listed by name, rather than computed on those that remain. For a large panel, such as a miRNA profile, choose the global mean instead.

  • Default: automatic, the targets marked as reference genes, else none
  • Scale to: the average by default; your control group, a sample or a calibrator

In full on the Analyze page

04In the app: Analyze › Statistics & apps

A test per target, saved as it ran

Pick the property that defines your groups, then the test. Tests run on the log of the normalized quantities, so the result is a fold change with its interval. A two-group test corrects its p-values across targets. ANOVA takes one target at a time, with the correction made over its pairs. The result keeps its settings.

  • Default: Welch’s t-test, α 0.05, Benjamini–Hochberg

In full on the Analyze page

Why it is done this way

Clarida follows the qBase model (Hellemans et al., 2007): each Cq becomes a quantity with the target’s own efficiency, as Pfaffl (2001) showed it should, and the error of each step is carried into the next. MIQE 2.0 asks for efficiency-corrected quantities reported with their uncertainty (Bustin et al., 2025).

One reference gene can shift with your treatment and take the fold change with it. The geometric mean of several stable ones is more robust (Vandesompele et al., 2002); for a panel of tens to hundreds of assays, the mean of all expressed targets works as well (Mestdagh et al., 2009).

Expression ratios are skewed: a doubling and a halving are the same size of effect. So tests run on log-transformed quantities, with your biological replicates as n, and the fold change comes back with an interval on the same scale.

What it doesn’t do

Absolute quantification, copy numberNot in Clarida: a standard curve sets the efficiency, not the amount
Choosing reference genes in this experimentIn a reference genes experiment, then mark the winners here
Two-way ANOVA, a test over several targets at onceOne-way ANOVA, one target at a time
Reading RNA-seq or microarray resultsNot in Clarida: you set them beside the fold changes yourself

Frequently asked questions

Do you need a standard curve?

No. Each target uses a standard curve in the experiment when there is one, then the efficiency in your library, then 2. Clarida shows which source each target used.

What if no target is marked as a reference gene?

Then there is no normalized result yet. Normalization lists your targets for you to tick the reference genes; for a large panel, switch it to the global mean.

What do the error bars show?

In the table, the standard error of each normalized quantity: how well it was measured. In a test, the 95% confidence interval of the fold change between your groups, from the biological replicates.

Does it work for miRNA and lncRNA?

Yes. The math does not depend on the kind of RNA. For a miRNA panel the global mean is often the better normalizer.

Can you use qPCR to confirm RNA-seq hits?

Yes: that is relative quantification of your candidates on reference genes shown to be stable in your samples, and the fold change with its interval is what you compare with the sequencing result.

Updated

Take last week’s run to fold changes.

Start from your run file, or open one of the demo experiments in your new workspace first. 4 experiments in your first month, then 2 a month. No card.

Other experiments

All kinds of experiment
  • Reference genes

    Validate a panel of candidates in one experiment: design it, run it, rank stability by geNorm M and V and by NormFinder, and save the ranking for your next normalizations.

    Start from
    Reference gene finder preset
    From a run file
    Yes, analysis only
  • Present or absent calls

    One call for every sample and target, made only when all replicates agree, with positive, negative and internal controls checked beside it and the cut-offs kept with the saved result.

    Start from
    Detection preset
    From a run file
    Yes, analysis only
  • Genotyping

    A genotype table from KASP, TaqMan, DMAS or melt assays: each marker declared once, each call with its quality, and known controls that catch swapped alleles.

    Start from
    Genotyping preset
    From a run file
    Yes, analysis only
  • Plate and instrument QC

    Fill a plate with one sample and one assay, run it, and get a pass or fail, the outliers by position and a heat map.

    Start from
    Plate homogeneity check preset
    From a run file
    No
  • Multi-plate studies

    Several plates as one study: sessions on different days, every run file matched, and each correction shown with its standard error.

    Start from
    Relative quantification preset
    From a run file
    Yes, analysis only