9 minutes ago5 min read
Reference Genes for qPCR: How to Choose and Validate Them (Beyond Just Using GAPDH)
9 minutes ago
5 min read
If your treatment raises GAPDH by half, every target you normalize to it will look a third lower, even if nothing about the target changed. That is less than one cycle of drift in the reference gene. The fix is simple: validate at least three candidate reference genes for qPCR in your own samples, and normalize to their geometric mean.
Few labs do. Across published gene-expression studies, the average number of reference genes used is 1.2, and only 15% of studies test a panel for stability before trusting it. With a single reference gene, a quarter of samples carry errors above 3-fold. Below is how to choose candidates, how to validate them, and what reviewers now expect you to report.
Why reference genes for qPCR cannot be assumed stable
Normalization works on one assumption: the reference gene's expression is identical in every sample, so any difference in its Cq reflects only how much RNA went into the reaction. If the treatment moves the reference gene, that movement is silently subtracted from every target.
Treatments do move them. Under hypoxia (1% O₂ for 24 hours), GAPDH mRNA rose between 21% and 75% across four cell lines, while 28S rRNA stayed flat. Worked through, that range is only 0.28 to 0.81 cycles of Cq drift, and it makes an unchanged target appear 17% to 43% downregulated. Nothing on the amplification plot looks wrong. The false result only appears in the ratio.
This does not make GAPDH a bad gene. In a model-based screen of colon cancer samples, GAPDH ranked among the three most stable candidates. The lesson is narrower and more useful: no gene is stable in every tissue under every condition, so stability has to be shown in yours.
How to choose candidate reference genes for qPCR
Start with more genes than you plan to use. Screening six to ten candidates gives the stability tools something to choose between. CLYTE's qRT-PCR protocol lists common starting points (ACTB, GAPDH, B2M, HPRT1, RPL13A).
Pick genes from different pathways. Two glycolytic enzymes, or two ribosomal proteins, tend to rise and fall together. A pairwise stability method will then rate them as mutually stable even if both are moving with your treatment.
Validate in samples that match the real experiment. Include every treatment group, time point and tissue you will compare. Candidates judged stable in untreated cells tell you nothing about treated ones, and published "best genes" vary between studies of the same tissue.
Validating stability with geNorm and NormFinder
Run both tools, and give them efficiency-corrected relative quantities rather than raw Cq values. An assay running at 90% efficiency and one at 100% will otherwise be ranked on a distortion of the tool's own making.
geNorm ranks stability and decides how many genes you need. Its stability value, M, is the average pairwise variation of each gene against all the others. Lower is better. Stable reference genes typically reach M below 0.5 in homogeneous samples such as one cell line, and below 1 in heterogeneous ones such as patient tissue. Treat the widely quoted cutoff of 1.5 as an outer limit only. geNorm then adds genes one at a time until the pairwise variation V between successive normalization factors drops below 0.15, starting from a minimum of three.
NormFinder checks whether a gene shifts between groups. geNorm has no notion of treatment groups, and it tends to favor co-regulated genes. NormFinder separates variation within groups from variation between them, which is exactly the variation that fakes a fold change. It needs more than eight samples to do this reliably.
When the two tools agree, you have your panel. When they disagree, trust NormFinder's between-group estimate for a treatment comparison, and drop any gene that moves with the treatment regardless of its overall rank.
Normalize to the geometric mean of the chosen genes. The geometric mean resists a single outlying gene and does not let a highly expressed gene dominate a weakly expressed one. It also dilutes, but does not erase, the damage from a gene that drifts: average GAPDH's 1.5-fold rise with two stable genes and the false downregulation shrinks from 33% to 13%. That is still enough to invent a finding, which is why validation comes before averaging, not instead of it.
If you want a second pair of eyes before committing to a panel, paste your candidates' Cq values by group into Soφ and ask which ones track the treatment.
The RefFinder shortcut, and why it misleads
RefFinder runs geNorm, NormFinder, BestKeeper and the comparative ΔCq method at once and merges them into one consensus ranking. It is convenient, and its input is the problem: it takes raw Cq values and ignores each assay's amplification efficiency, and its BestKeeper ranking uses a different statistic from the original software. When efficiencies differ, its rankings shift. Treat it as a cross-check, never as the verdict.
What to report under MIQE 2.0
The 2025 revision of the MIQE guidelines asks you to justify both the choice and the number of reference genes, to show their stability under your specific tissue and conditions, and to state the validation method and the normalization calculation. In practice, that means a supplementary table listing every candidate tested, its stability values from each tool, and the genes you kept.
The best reference genes for qPCR are the ones you validated yourself
GAPDH, ACTB and 18S are reasonable places to start and unreliable places to stop. The reference genes that belong in your paper are the ones that held still across your own treatment groups, chosen from different pathways, confirmed by more than one method, and combined by geometric mean. Half a cycle of drift is all it takes to report a change that never happened.
Running the qPCR itself?
Frequently asked questions
Is GAPDH a good reference gene for qPCR? Sometimes. It is stable in some tissues and shifts with others, notably under hypoxia. Validate it alongside other candidates rather than assuming either way.
How many reference genes should I use? At least three. geNorm tells you whether a fourth adds anything: stop adding genes once the pairwise variation falls below 0.15.
What is a good geNorm M value? Below 0.5 for homogeneous samples and below 1 for heterogeneous ones. The often-quoted 1.5 is only an outer limit.
Should I use geNorm or NormFinder? Both. geNorm sets how many genes to use; NormFinder catches genes that shift between your treatment groups, provided you have more than eight samples.
References
Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A, Speleman F. Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biology, 2002;3(7):research0034. DOI
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 Research, 2004;64(15):5245–5250. PubMed
Hellemans J, Mortier G, De Paepe A, Speleman F, Vandesompele J. qBase relative quantification framework and software for management and automated analysis of real-time quantitative PCR data. Genome Biology, 2007;8(2):R19. DOI
Chapman JR, Waldenström J. With reference to reference genes: a systematic review of endogenous controls in gene expression studies. PLoS One, 2015;10(11):e0141853. DOI
De Spiegelaere W, Dern-Wieloch J, Weigel R, et al. Reference gene validation for RT-qPCR, a note on different available software packages. PLoS One, 2015;10(3):e0122515. DOI
Zhong H, Simons JW. Direct comparison of GAPDH, beta-actin, cyclophilin, and 28S rRNA as internal standards for quantifying RNA levels under hypoxia. Biochemical and Biophysical Research Communications, 1999;259(3):523–526. PubMed
Bustin SA, Benes V, Garson JA, et al. The MIQE guidelines: minimum information for publication of quantitative real-time PCR experiments. Clinical Chemistry, 2009;55(4):611–622. PubMed
Bustin SA, Ruijter JM, van den Hoff MJB, et al. MIQE 2.0: revision of the minimum information for publication of quantitative real-time PCR experiments guidelines. Clinical Chemistry, 2025;71(6):634–651. PubMed
