Is this the geNorm algorithm from the 2002 paper?
Yes, with one change, and its M values checked against qbase+.
| The method | Vandesompele et al., 2002, Genome Biology: the stability measure M and the pairwise variation V |
|---|---|
| From | Its first author, Jo Vandesompele, advises Clarida; the paper has been cited more than 20,000 times |
| One change | The last two genes get an M value each, as in qbase+, so the ranking goes down to a single gene; in 2002 the pair shared its value |
| Checked against | qbase+: the same M values on a published example |
| What goes in | Cq (Ct) values, converted with each gene’s own efficiency; the original Excel applet took relative quantities |
How it works
01In the tool: geNorm › Data
Paste your Cq values as a table
One column per gene, one row per sample, with technical replicates as repeated names. Paste it from a spreadsheet; an Excel, CSV or text file loads too. If your export lists one well per row, arrange it first. From 3 to 100 genes, on at least 3 samples.
- Efficiency per gene: 100% unless you set it, as E or a percentage
- Replicates: averaged as the mean, or the median; flagged when more than 0.5 cycles apart
02In the tool: geNorm › Data › Recalculate
Leave out a sample without deleting it
Switch a failed sample or a doubtful gene off in the grid, then press Recalculate. What you pasted stays as it was, so switching it back on gives the first ranking again.
03In the tool: geNorm › Results › Stability ranking
See which genes hold steady
The least stable gene is left out step by step, and each time M is worked out again, until one gene is left. As a guide, M below 0.5 suits a homogeneous set of samples and below 1.0 a mixed one (Hellemans et al., 2007). A recommendation above the chart names the genes to use.
04In the tool: geNorm › Results › Optimal gene count
Know how many genes to normalize with
The pairwise variation V compares the normalization factor of n genes with that of n + 1. Below 0.15, one more gene adds little, so n genes are enough. The PDF report, with the ranking and this chart, comes by email.
What geNorm doesn’t do
| Use your sample groups | NormFinder does, on the same data, one tab over |
|---|---|
| Download the PDF report | It comes by email; see an example |
| Keep the ranking with your run | Reference genes in Clarida, inside an experiment |
Frequently asked questions
Is geNorm free?
Yes, with no account. Saving a ranking to your account needs a free one, and the PDF report comes by email.
How many reference genes should you use?
At least two, and geNorm tells you how many: when the pairwise variation V between n and n + 1 genes falls below 0.15, n genes are enough.
What is a good geNorm M value?
Below 0.5 for a homogeneous set of samples, below 1.0 for a heterogeneous one (Hellemans et al., 2007). These are guides: geNorm ranks genes among themselves rather than against a cutoff.
Why do the top two genes have different M values?
Because the ranking goes one step further than the 2002 method. There, the last two genes are each other’s only comparison, so they share one M value and stay a pair. Clarida gives each its own value, as qbase+ does, which matters when you need one most stable gene or want to trim a panel. Normalizing with at least two genes keeps both anyway.
Should you use more than one algorithm?
No. One reliable algorithm is enough: where geNorm and NormFinder disagree, the genes in question can all be good choices. Clarida shows both rankings side by side if you want to compare.
How does geNorm compare to NormFinder?
geNorm ranks genes by how much each pair varies across all samples and ignores groups. NormFinder models the variation within and between your sample groups, so it can catch a gene that shifts with your treatment. Their rankings may differ when your samples fall into distinct groups.
How does geNorm compare to BestKeeper?
BestKeeper (Pfaffl et al., 2004) works on raw Cq values with standard deviations and Pearson correlation, without efficiency correction, and says nothing about how many genes to use. Clarida does not offer it.
How does geNorm compare to the comparative ΔCt method?
Both compare candidate genes in pairs. The ΔCt method (Silver et al., 2006) takes the standard deviation of Cq differences and fits in a spreadsheet; it gives no number of genes to use. Both can favor genes that are regulated together.
