What you get
One experiment holds every plate of the study. Each target’s Cq values are corrected for the shift between runs, then averaged, normalized and tested as if every sample had been on one plate.
The correction is shown, not hidden. A matrix lists each target against every run, with its calibration factor (CF) in cycles and the standard error; click a cell to see how far each shared sample sits from it. A run that shares no sample with the others is marked uncalibrated, with a correction of 0. Add the matrix and its settings to a PDF or HTML file with the Report builder.
Is it qbase+’s inter-run calibration?
The idea is the one from the qBase paper (Hellemans et al., 2007), by the people behind qbase+: samples measured on more than one plate tell you how far each run is off. Where Clarida applies it differently from qbase+:
| Which samples calibrate the runs | Every sample measured for a target on two or more runs, except negative controls and excluded replicates. You mark no calibrators and rename no samples per run |
|---|---|
| What is corrected | Each target’s Cq, per run, before replicates are averaged and normalized; qbase+ calibrates the normalized quantities |
| The same calibrators on every plate | Not needed: one fit per target over all of them, so plates linked through a chain are calibrated together |
| The same sample on each run | The same cDNA, measured again. A separate cDNA synthesis of the same RNA is a different sample: don’t give it the same name, or its own offset is read as the run’s |
| A sample typed as calibrator | Used by Scale to calibrator in Normalization, to set the reference level; it does not link plates |
How it works in Clarida
01In the app: Design › Run Layout
Plan the plates, and choose what may be split
When the reactions do not fit on one plate, Clarida lays them out over as many as needed. You decide whether a block, grouped by assay or by sample, may be split over two runs. Left on automatic, blocks stay whole when everything fits on one plate and split when it does not, with the reason shown under the setting.
- Plate crossing: keep blocks whole, or split across runs; automatic by default
Read the method: Plate layout generation
02In the app: Execute › Sessions
Run the plates in sessions, on different days
Group plates run on separate days, or by other people, into sessions. Collect Material and Prepare Mixes then list only what the active session needs. Each item shows any other session that uses it too (“Also needed in”). For a mix whose wells span several runs, a toggle in Design › Mix decides whether it is made once for all of them or per run.
- Mix prep for a split group: Auto by default (follows the sessions), per run, cross-run
In full on the Execute page
03In the app: Import › Import & annotate › Link
Each run file linked to the plate it came from
Drop in several run files at once. Each is linked to its planned run, or kept as a new one, so the measurements land on the right wells. Give each biological sample one name across all plates: that shared name lets it calibrate the runs it appears on.
In full on the Import page
04In the app: Analyze › Cq processing › IRC
Run-to-run shifts measured and removed
The IRC part appears as soon as a sample is measured for the same target on two or more runs. For each target, Clarida fits one shift per run from all its calibrators together and subtracts it from that plate’s Cq values. A run with no calibrator in common with the rest keeps a shift of 0, with a warning.
- Inter-run calibration mode: apply (the default), skip, or external (the calibrated Cq from the file, offered only when it has one)
- Standard error limit: 0.3 cycles
In full on the Analyze page
05In the app: Analyze › Cq processing › Normalization
Normalized over all plates at once
After calibration, the replicates of a sample from different plates are pooled, and normalization runs on the whole study. Reference genes can sit on plates of their own: each target is calibrated separately.
- Method: automatic by default, reference genes when any are marked
In full on the Analyze page
Why it is done this way
A run adds its own small offset to every Cq it measures: from the plate, the mix aliquot, the day. Within one run it cancels when you compare samples. Across runs it does not, and it can look like biology: the qBase paper showed the same samples giving systematically higher Cq values on one plate than on the other.
The same paper gives two remedies. The preferred one is sample maximization: everything measured for a given gene on one plate, so nothing between runs needs correcting. Clarida lays the study out that way with blocks by assay, kept whole; automatic layout does not choose it once several plates are needed, so set it yourself. When samples must be spread over runs, identical ones measured on each plate give a calibration factor, and several give a more precise estimate than one.
Clarida fits that idea to any layout. For each target it solves one additive model (levels of the shared samples plus run shifts) by least squares, so calibrators need not be the same on every plate. Every factor comes with its standard error and each calibrator with its residual, in place of one quality score. A large factor, a wide error or one stray sample points at the run to check. MIQE 2.0 asks for the calibration between runs to be reported (Bustin et al., 2025).
What it doesn’t do
| Calibrators placed on every plate for you | Add them yourself: the generator places a sample once per assay |
|---|---|
| Plates from two separate experiments | Not calibrated against each other: the correction works within one experiment |
| Down-weighting a wild calibrator | Flagged, not down-weighted; exclude its wells and the factors are fitted again |
| Calibration for some targets only | On or off for the whole experiment |
Frequently asked questions
How do you mark your inter-run calibrators?
You don’t. Give each biological sample one name across all plates. Once measured for a target on two or more runs, it links them. Negative controls never count.
What if a plate shares no sample with the others?
That run gets a correction of 0 for the target, and the matrix marks it uncalibrated with a warning. If the plates form two groups with no sample in common, you are told the two are not calibrated against each other.
Your file already has calibrated Cq values. Can you keep them?
Yes. Choose External, and Clarida keeps the calibrated Cq from the imported file, with no correction of its own. The option is offered only when the file carries such values.
Do reference genes have to share a plate with the targets you study?
No. Each target is calibrated on its own, and normalization combines them afterwards. What matters is that each target’s samples sit on one plate, or are linked by a shared one.