If I understand your question, you are in part inquiring whether a Ct value (basically a log base 2 quantity) is a better starting point than relative fold expression (a ratio). I am not a fan of ratios. Raw ratios have unfortunate distributional properties for many statistical tests, and they have no absolute meaning. Interpteting "Jim lost 20 times more weight than Fred" depends on how much Fred lost. Without that knowledge, the difference between them could amount to a few micrograms, and the "20 times" statement loses biological meaning. Do you really care about a gene that is not really expressed in one condition and not expressed at all in another? Maybe not.

thanks Tamotsu

Why you want to do 3 Factorial CRD. What's the purpose. Please explain

Since you are interested in gen expression levels, use the relative fold expression values.

You could in principle use the Ct values too, but then you should take into consideration the Error Propagation, which would make your calculations damn complicated.

Thanks Peter, I was thinking the same...

Vajinder, my work is on differential mineral accumulation in plants, I want to see the interaction between tissue X gene X genotype etc, dats why CRD...

If I understand your question, you are in part inquiring whether a Ct value (basically a log base 2 quantity) is a better starting point than relative fold expression (a ratio). I am not a fan of ratios. Raw ratios have unfortunate distributional properties for many statistical tests, and they have no absolute meaning. Interpteting "Jim lost 20 times more weight than Fred" depends on how much Fred lost. Without that knowledge, the difference between them could amount to a few micrograms, and the "20 times" statement loses biological meaning. Do you really care about a gene that is not really expressed in one condition and not expressed at all in another? Maybe not.

thanks sir, I get your point and what do u think about plotting graphs, which will give a better representation fold expression or Ct values?

Personally, I would graph Ct values as boxplots for each group and see how it looks.

sir, I have never used boxplots, will you please guide me, i mean what tool I should use to draw them, I have Origin Pro6.1, will I b able to draw them with this software??

And why the boxplots are better representation?

Fold expression value can be taken

Thomas, I understand what you argue about with respect to fold changes, but I partly disaggree.

With your example, fold change looks are much more like "Fred lost 5% of his body weight on his new diet while Jim did not change significantly. In contrast with this, in the last year's experiment Jim lost 7% during the tennis course while Fred lost only 2%"

What are Ct values? Number of cycles in a PCR reaction of yet unknown (or not yet calculated) yield (doubling probability) to reach the same light intensity coming from (unknown amount of) amplified DNA (of perhaps different lengths). This is by far not a biological measure. This is a raw data, that enables us to calculate if a gene is overexpressed under given conditions relative to control conditions. Following your example, Ct is like "Jim went 10 more times to the pub than Fred, and he drank usually an unknown Y% as much and now they have the same size of kidney." :)

Hi Niloofar,

If you want to have the fold of over-expression/down-regulation of your target gene in your sample group compared with your control group you have to use the final number(2*-ddCt).in this situation you can not have Error Bar for your Control group,if you want it, use number of 2step before(dCt). you can analyze your numbers by SPSS, Excel, or GraphPad Prism simply.

I appreciate if i'm wrong others notice.

thanks Peter n Mansoureh

Hi Peter,

Ratios are problem with genes in a way they are not a problem things like weight, blood pressure, etc. The latter measure things we have an intuitive feeling for. If my systolic blood pressure rises by 10%, most people could guess that my bp rose by about 12 points, and they know, clinically, that this is not great, but not a cause to run to the emergency room. If my core body temperature rises by 10%, it's probably too late to do anything but arrange for a funeral. It's not clear with low expression genes whether a 10% difference matters. I've never heard of a situation where it does..

Any statistical analysis should be done with Cts (or dCts if you like) due to the issues Thomas raised. The statistical analysis will yield estimates for the mean Ct of each treatment group. The differences between those can be plotted in approximately fold change units by taking 2^(estimated mean Ct). Error bars are calculated in units of Ct and the ends of those lines can be shown in fold change as well (e.g. 2^(estimated mean Ct + 95% ci)).

Thomas and Michael, thank u very much!! :)

I am comparing the gene expression levels in different tissues (root, stem, leaf) of two different varieties (v1 & v2) of a plant through qPCR, keeping the root as calibrator. For analyzing my data, should I keep the root of v1 as calibrator for either varieties or do I need to take separate calibrators i.e. root of respective varieties as their individual calibrator.

What program can I use for microarray analyse when I used agilent array, but my scan is perkerHelmer?

I would derive the mean and 95% confidence intervals of the ddCt values,

Then transform these to fold changes, and plot.

The below reference recommends doing all the statistical work on the ddCt values

(or two sets of dCt values prefereably). Only change to fold-changes after everything is done.

1. Yuan JS, Reed A, Chen F, Stewart CN. Statistical analysis of real-time PCR data. Bmc Bioinformatics. 2006;7.

Thanks Colin