Fold change significance - (Jun/18/2012 )
Hi, so i have performed QPCR on some tissue (induced and control). I had 5 different dogs from which i took tissue. I then grew the cells to confluence, and set up an experiment either induced or normal/control in triplicate on three seperate days.
I had 2 reference genes of which i found the geometric mean, and then i worked out for each sample
Delta CT (CT gene of interest - CT geomean ref gene)
Delta Delta CT (Delta Ct induced - Delta CT normal)
and fold change (2^-DeltaDelta CT)
i averaged the induced and normal fold change at the very end, to get the averga fold change for each gene under the induced and normal condition from which i used to draw graphs etc.
My problem now is how to i tell if the fold change is significant or not - which significant test should i use?
I would assume that the data is not normally distributed so it would have to be a non-parametric test, and i was assuming that the Man Whitney U test would be appropiate, but when i try and run this in minitab on the fold change values for a gene, it will not run, as the fold change for the control is 1 for every sample - could i therefore run the test on the CT values or DeltaDelta CT values instead?
If anyone could give me any help on how to assess significance of the data (i have minitab and graphpad prism as stats/graphing programs) then it would be much appreciated.
for biological group comparisons a t-test would be suitable. this test is also used in the ABI Data Assist software. Fold-change values are often log2 transformed for statistical evaluation.
I have been dealing with the same question for a long time. The key point is that there are two levels of statistical analysis involved: (A) Within each group (technical replicates), ( Between groups (when you compare the biological replicates).
1) At the technical replicate level, make sure the ddCt numbers of your experimental sample is significantly different that the ddcts of the control group. This is done by a simple t-test. If they are significant, then just get the mean fold change and use it for the rest of the analysis.
2) Do the step 1 for each of biological replicate analysis. You will end up having 3 numbers now. Let's say biological replicate #1 gave you 1.5 fold change, #2 gave you 2.0 fold change and #3 gave you 3.1 fold change. Since fold change means compared to that of control, it assumes that all the controls are 1.
3) Usually, you don't have to show the control group in your data presentation, because everyone knows that they are all 1, with no variation (no error bar). Look at this data:
As you see, they don't include their control. They could actually have a bar for control. But that bar wouldn't have any error bar. Sometimes people draw a dotted line, crossing the Y-axis. To show the threshold of control. See this:
4) Finally, If you want to compare two groups only (group 1 vs. group 2) a t-test is the right test. If you have more than two groups and you want to show that on a graph, then you have to run ANOVA (you will find this anywhere on the web) with a post-hoc test. ANOVA would tell you if your groups are different in general. The post-hoc test runs a paired comparison to identify which specific pairs are different.
I'm not completely sure about what I said. This only what I have learned form the literature.