By the way, always log transform your fold changes or relative quantities before you apply statistics. More details in "qPCR data analysis – unlocking the secret to successful results" by Jan Hellemans and Jo Vandesompele, book chapter in "PCR Troubleshooting and Optimization: The Essential Guide Caister Academic Press 2011, ISBN 978-1-904455-72-1". After login to a (free) MyBiogazelle account, you can download the full text.

http://www.biogazelle.com

Mirko, I understand your question, but I do not agree to the pretext that your calibrator has no standard deviation. You should incorporate the error associated with your mean Ct values for the calibrator and test samples in your calculation . The standard deviation calculations follow the standard laws of error propagation. Attached find an example excel sheet, the calculations start after you have calculated the normalized expression ratio using the Pffafl method. After you have incorporated the error in your calibrator and test values, you can proceed with a T test.

Dear Amey,

Thank you for your complete answer, but my question begins a bit earlier. Pfaffl method (or efficiency method) uses Cq average to calculate ΔCq mean. Then, ΔCq mean is used to estimate the fold of change in transcript amount. Afterwards, this change in transcript amount is normalized dividing by the fold of change in transcript amount of an internal control (reference gene). So, the only part where I have error is in the calculation of Cq average which has standard deviation asociated and I don't know how to transfer that error to the final result.

I added a new sheet with calculation and a new example.

Thank you again! I hope you can help me

Mirko--

You could use one of your sample as calibrator and calculate the fold change for the other biological replicates (for both control and test groups) and generate an SD. I have always had problems with this method (to use one of the samples as calibrator), we routinly a standard RNA (pools from multiple samples/tissues) as a reference and we calculate the fold change for all the samples. Using this has an adevantage as the same RNA prep is used for every experiment (and expect the same CT in every run), our inter experimental variability values are quite reliable and we get to know if our PCR has gone wrong at any point.

Hope this helps.

Best luck

Hi Mirko I looked at your excel sheet calculations, and there are some mistakes, firstly you have applied the Pffaffl method wrongly. This method is meant for comparing two genes with different amplification efficiencies not for comparing two different amplification efficiencies for the same gene in different samples. The E reference is the amplification efficiency of the reference gene, and Etarget is the amplification efficiency of the target gene. You can apply the Pffafl method for different amplification efficiencies for the same gene, but not in the way you have done.

Secondly you leave the standard deviation of the Cq cycles unaccounted for. I think you have also somehow mixed up the labelling, I think you meant biological replicates and not technical replicates right?

Amey... we normally apply stats to biologic replicates and not to technical replicates rite ??

Yes Deepak, we apply statistics for biological replicates, but the error from the technical replicas has to be propagated into the final results.

Thank you again Amey!

As I understand, Pfaffl method is not exclusive to compare to genes with diferent amplification efficiency, but you can also use it to estimate the relative transcript amount between a reference condition (calibrater) and a test condition using Eq1 (see attachement). By the other hand, I did not mix up the labels and, yes, I left unaccounted Cq standar deviation because this is my main question, How can I transfer calibrater Cq standar deviation to the final result using Pfaffl method?

I'm using an Eco Illumina to carry out qPCR reactions and the Eco software estimates RQ as I did. By the other hand, I looked for papers where use Pfaffl method and almost no one consider error over calibrater or doesn't say how they did to estimate it.

Anyway, tired to figure out this problem I use -ddCq method instead (see link to reference). They describe very well how to estimate calibrater error.

http://www.sciencedirect.com/science/article/pii/S1046202301912629

Respecto to error propagation, I did as you suggest with a little modification described in attached letter (see link). This formula considers bias caused by normalization against calibrater if the ratio is too high (or too low).

I also attached a data sheet, with calculus for RQ and error progation for -ddCq method if anybody is interested.

Thank you very mucho both of you!

your answers were really helpful to me

Regards

Mirko.

http://www.plantcell.org/content/21/4/1031.full

Here is Eq1

And here is the macro!

Apply SPSS----->analyse------> T-test

You may also want to have a look at Willems et al., 2008. This strategy will work in your case if you have performed 3 independent experiments. If you have measured your calibrator samples in the same experiment, then you should use just one of them (irrelevant which one) as a calibrator sample. Biogazelle's qbase+ software uses the average expression level across all samples as a kind of virtual calibrator sample (more details in Hellemans et al., Genome Biology, 2007).

Article Standardization of real-time PCR gene expression data from i...

By the way, always log transform your fold changes or relative quantities before you apply statistics. More details in "qPCR data analysis – unlocking the secret to successful results" by Jan Hellemans and Jo Vandesompele, book chapter in "PCR Troubleshooting and Optimization: The Essential Guide Caister Academic Press 2011, ISBN 978-1-904455-72-1". After login to a (free) MyBiogazelle account, you can download the full text.

http://www.biogazelle.com