Skip to content

Free tools

NormFinder: rank reference genes by sample group

Rank candidate reference genes by their variation within and between your sample groups. Checked against the authors’ Excel add-in.

You start withCq values, and your sample groups if any
You end withBest gene and pair, geNorm beside it
Samples in two groups, the stability ranking under them; then how far each gene shifts between the groups, and geNorm’s ranks beside NormFinder’s.
On this page

Does it match the NormFinder add-in?

Yes: on the authors’ own example, to within 1e-9 once the units match.

The methodAndersen, Jensen and Ørntoft, 2004, Cancer Research: a model of the variation within and between sample groups
Written fromThe published model, not from existing code
Checked againstThe authors’ Excel add-in on their bladder cancer example: the same ranking, with stability values, variances, group differences and the best pair within 1e-9
Unitslog2, the Cq scale, as geNorm’s M; the add-in reports natural logs, 0.693 times these
What goes inCq (Ct) values, not the linear quantities the add-in asks for, converted with each gene’s efficiency

How it works

01In the tool: NormFinder › Data › Assign groups · Paste groups

Give every sample its group

Paste one column per gene and one row per sample, Cq values as your cycler gives them. Groups come three ways: a column in your file headed Group, Condition, Treatment or Class; samples you tick in the grid; a list pasted in the order of your data. Each group needs at least 2 samples, and the tool says when one has fewer than 8, the size at which NormFinder is most reliable.

  • Efficiency per gene: 100% unless you set it, as E or a percentage
  • Candidate genes: from 3; the tool warns below 5, as the authors advise 5 to 10

02In the tool: NormFinder › Results › Stability ranking

See which gene stays put across conditions

Each gene’s stability value combines how much it varies within your groups with how far it shifts between them; lower is more stable. A table gives each value beside the pooled within-group SD and the group difference. There is no cutoff, so compare genes by rank.

03In the tool: NormFinder › Results › Group differences

Spot the gene that moves with your treatment

With two groups or more, a chart shows how far each gene shifts between them, and the spread inside each one. A gene can vary little within groups and still move with the condition: this is where it shows.

04In the tool: NormFinder › Results › Best pairs

Pick a pair, not only a gene

Two genes whose shifts cancel out can make a better pair than the two best single genes, so with four candidates or more the best pair is not always the top two. The ten best pairs come with their combined stability.

05In the tool: NormFinder › Compare methods

Read geNorm beside it

The same analysis runs geNorm. Compare methods lists both ranks per gene and marks the genes where the methods disagree. There is no combined rank: a disagreement is worth a look, not an error to average away.

What NormFinder doesn’t do

Say how many genes to usegeNorm’s pairwise variation does, on the same data
Run while only some samples have a groupNot until every sample has one, or none does; the tool lists the samples still missing
Keep the ranking with your runReference genes in Clarida, inside an experiment

Frequently asked questions

Why do the values differ from the NormFinder Excel add-in?

The units. Clarida reports NormFinder values in log2, the Cq scale; the add-in reports natural logs, which are 0.693 times these. The ranking is identical, and on the authors’ bladder cancer example the values match to within 1e-9 after converting.

Is NormFinder free?

Yes, with no account. Saving a ranking to your account needs a free one; the PDF report, with both methods, comes by email (see an example).

Does it work on a Mac, without Excel?

Yes. It opens in your browser with nothing to install, whatever your operating system or Excel version.

Do you need sample groups?

No, but they are the reason to use NormFinder. Without groups it treats all samples as one set and cannot tell a shift between conditions from ordinary spread. geNorm does not use groups and runs either way.

What is a good NormFinder stability value?

There is no cutoff. NormFinder’s authors say the value does not tell you whether a gene is acceptable, so compare genes by rank. geNorm’s M thresholds do not apply either: the two share a unit, not a meaning.

What happens with missing values?

Both methods need a complete table, so the tool removes samples or genes one at a time, each time the one that clears the most gaps for the fewest measured values lost. It keeps at least 3 genes, 3 samples and 2 samples per group, and names what it left out in the grid. A gap in one technical replicate is fine: the others are averaged.

How does NormFinder compare to geNorm?

geNorm ranks genes by their pairwise variation across all samples and ignores groups; NormFinder models the variation within and between your groups. Their rankings can differ when your samples fall into distinct groups. The geNorm page covers whether you need more than one method.

Cite this tool

Format

Clarida. (2026). Reference gene finder (Version 2026.11.0) [Web application]. Retrieved November 18, 2026, from https://app.clarida.bio/reference-gene-finder/normfinder

Also cite the method

  • NormFinder: Andersen CL, Jensen JL, Ørntoft TF. Normalization of real-time quantitative reverse transcription-PCR data: a model-based variance estimation approach to identify genes suited for normalization, applied to bladder and colon cancer data sets. Cancer Res. 2004;64(15):5245–50. doi:10.1158/0008-5472.CAN-04-0496

Updated

Find out whether your genes shift between groups.

Paste your Cq values with a group for each sample; the best gene and pair come back, geNorm beside them.

More free tools

All free tools
  • geNorm reference gene finder

    Rank candidate reference genes by stability, and see how many you need.

    You start with
    Cq (Ct) values of your candidate genes
    You end with
    Ranking, how many genes, PDF by email
  • Plate homogeneity

    Test whether every well on your instrument reads the same.

    You start with
    Cq values of a uniform 96-well plate
    You end with
    Pass or fail, outlier wells, heatmap
  • Plate editor

    Plan each well’s sample and assay, then paste the layout into your instrument software.

    You start with
    Sample and assay names, or a run file
    You end with
    A plate layout to copy or export