# inter-run calibration - (Jan/24/2011 )

Hi,
I need some help in understanding some formulas...
To begin at the start, here my setup:
LightCycler
3 biological replicates for:
1. control
2. treated with substance a
3. treated with substance b

1 calibrator which is a mixture of former test runs of the same cell culture

every biological replicate/sample and also the calibrator were measured in triplicates for:
3 reference genes
6 genes of interest

per run I just measured one gene, and for space reasons only 1/3 to 2/3 of the samples.

To do the inter-run calibration I tried a formula I found here (sorry, do not know how to insert the formula here, if possible): page 12 formula 13' (not 13) in the pdf
http://genomebiology.com/2007/8/2/R19

Well, I think it means:
the c-th-root of the product of all calibrator-samples for one gene in one run, where c is the number of used calibrators.
So far, but:
- If this is so, where is the correlation between runs??? I mean it's just the ... product of all calibrator samples in one run!????
- If I am wrong and it should mean, the c-th root of the product of all calibrator-samples for one gene in all runs. It doesn't make sense either, because you just get the same calibration factor for all runs!?????

I hope I could explain what I do not understand. Some ideas or other ways how to do the inter-run calibration?
Ah, please no software I have to pay for!

-ThethirdD-

Hi, the result of the formula is the geometric mean of your NRQs from the IRCs of run l. This value (CF) is later used in formula 15' to calculate the calibrated normalized relative quantities from run l (CNRQ). you have to do this for every run with the respective CF of the IRCs from the same run.

I hope this was helpful, its some time ago when I crawled through these formulas

-tea-test-

tea-test on Mon Jan 24 17:16:22 2011 said:

Hi, the result of the formula is the geometric mean of your NRQs from the IRCs of run l.

-> This is my problem, this is what I do not understand!!! How can you link different runs, if you just take the geometric mean of samples within one run????
So in concrete, that's what I understood. I take the geometric mean of the IRCs of run 1 and have calculated the CF for run 1. Then I do the same for run 2, 3 and so on. What I get is one CF for every single run.

This value (CF) is later used in formula 15' to calculate the calibrated normalized relative quantities from run l (CNRQ). you have to do this for every run with the respective CF of the IRCs from the same run.

-> Yes, ok, that's what I also thought what the formulas say, but: I take the CFs which I calculated for every single run and take the quotient of the sample NRQs and the run-specific CF. Fine, this is what I do not understand. Where is the link between the runs???

Or, maybe, is the idea behind it that in perfect runs the IRC would have the same Cqs and efficiencies and therefore the CF should be exactly the same for every run? And if not the CF differs between runs and this is the "link" between the runs which does the calibration???????

I hope this was helpful, its some time ago when I crawled through these formulas

Yes, thanks. Maybe you set the ball rolling.

-ThethirdD-

hi, the link between the plates/runs is that you are using the same IRC on all plates for the calibration. So, you are calibrating your samples for every single run by using another sample or more than one (=IRC) that is present on all plates of your experiment. Therefore, you can correct for the technical variance between the different runs.

-tea-test-

sorry, for some reason I do not get email notification on replies...

Thanks for the answer, I think I understand now.

-ThethirdD-