Hello,
I have done RT-PCR in which I compare the expression of a gene of interest (TGF-beta1) in a control group with 4 samples and a treatment group with 4 samples.
I calculated delta CT for the samples as follows:
delta CT = CT (GAPDH) - CT (TGF-beta1)
I calculated the average of the delta CT in the control group and named it: delta CT (control group)
And I calculated the average of the delta CT in the treatment group and named it: delta CT (treatment group)
I calculated delta delta CT as follows:
delta delta CT = delta CT (treatment group) - delta CT (control group)
which allows me to calculate the fold changes in gene expression.
However, my question remains which test I should ideally use, based on my delta CT values, to calculate a p-value and to determine whether my results are statistically significant or not. I want to test whether my gene of interest is significantly higher expressed or lower expressed in my treatment group. Since I think my data do not follow a normal distribution, I cannot use the unpaired T test for example. For a Wilcoxon-Mann-Whitney test, my sample size (4 samples in both groups) might be too small.
Any help would be appreciated.
Kind regards,
Maxim
I have never seen any publication demponstrating that the assumption of normally distributed dCt values would be implausible (if someone has such a publication I would be grateful to learn!).
The t-test is prefectly fine to test the null hypothesis that the expected difference in the dCt between the treated and the control group (= the expected ddCt value) is zero - and this is exactly what you want to know. Note that the WMW test does not test the same hypothesis. That test is about the stochastic equality, the null hypothesis is that Pr(dCttreat > dCtctrl) = Pr(dCttreat > dCtctrl) = 0.5. With n=4 in each group the smallest p-value the WMW test can give you is 0.02857. This might be small enough. But if you "should" use the t-test or the WMW test depends on the hypothesis you like to test. If you want to test the expected difference and the assumptions of the t-test would be definitifely implausible for the dCt values, then the only way out is to bootstrap the sampling distribution - but this is not very fruitful for small samples.
What is more important than the test is that the assay is well optimized (i.e. you get robust and optimum amplification efficiencies and you do not amplify unspecific products) and that GAPDH is a valid reference gene for your treatment (it must not be regulated in response to the treatment; e.g. if the treatment invoves hypoxia then it is likely that GAPDH will be itself regulated and thus won't be a valid reference gene).
Thank you for your reaction.
I have read your reactions on other questions related to the statistical analysis after quantitative PCR. Optimally I would like to illustrate my article with a graph wherein I show the delta CT for the gene of interest in the treatment group compared to the delta CT for the gene of interest in the control group, because it is shows indeed the upward regulation and downward regulation in a visually more equal way than the fold change. For example: http://d3n1qfjutz9qb9.cloudfront.net/content/ajpgi/289/6/G1036/F2.large.jpg. I should show the difference of genetic expression whereby the expression in the control group is 1 on a logarithmic scale as illustrated in the example.
If, for example, the delta CT in the control group is -5 and the delta CT in the treatment group is -12, can I rewrite that: for an expression of 1 in the control group on a logarithmic scale, the expression in the treatment group is -6 (because -12 plus 6 equals -6)?
Don't complicate things.
It seem that what you actually want to show are ddCt values, not dCt values.
Then simply show the ddCt values. That is very simple and direct. In fact positive ddCt value indicate upregulation and negative ddCt values indicate downregulation (the way you calculated them). You may show these ddCt values as a bar garph (although this is not my favorite).
It is NOT recommended to show dCT values as bargraphs, because there is no defined reference value of dCt values. Your idea to place the horizontal axis just at the dCt value of the control group and thus let the bar heights indicate the difference of the dCt values between the treated groups to the control group is actually nothing else but showing the ddCt value, just in a bit indirect and possibly confusing way. This "trick" would work for one target gene, but not for several different target genes, where you would have to chose a different value for the intercept for each gene, what is not possible within on diagram.
So if you want to show dCt values then please use a graph like a boxplot, a beeswarm plot or a dotplot with error bars (see link), but not a barplot.
https://www.researchgate.net/file.PostFileLoader.html?id=5357e630d2fd64032d8b457d&assetKey=AS%3A273518703382529%401442223348718
- Badges
- Science topic