Real time PCR normalization algorithm

When performing normalization of real time PCR results, I found two ways of doing it:

1. 
   

   In my lab they follow the next layout: $\text{Efficiency}^{-(CT\ _{\large\text{interest gene}} - CT _{\large\text{ housekeeping}})}$ each time controls with controls and treatments with treatments. Then they divide the all values (treated and controls) each with controls.

   
   


2. 
   
   

   On the other hand I found another way to normalize that follows this steps: $\text{Efficiency}^{(CT_{\large\text{control}} - CT_{\large\text{treated}})}$. Then you make the same for the normalizing gene and divide the first by the last.

   
   

The values obtained are similar but not exact. Also I noticed that first algorithm gives rather different values for most diluted when performing a standard curve. Which is correct?

Answer

First step is the calculation of efficiency, denoted by lets say $E_{gene}$. See this post for calculation of primer efficiency.

So the fold change for that gene will be calculated by $E_{gene}^{-\Delta Ct_{gene}}$

Where:

$\Delta Ct = Ct^{treated} -Ct^{control}$

But these Ct values are not normalized. For normalization, you take some reference gene which need not be a housekeeping gene. Reference gene should chosen such that it is not affected by the treatment. In some cases usual housekeeping genes can also get affected for e.g. in treatments affecting cell cycle or differentiation. In such cases you should use a spike-in (an artificial RNA bearing low resemblance with any known gene).

You can either normalize the fold change (Case A) or find fold change of the normalized expression (Case B).

Means the same mathematically.

Case A:

$E_{gene}^{-\Delta Ct_{gene}} / E_{ref}^{-\Delta Ct_{ref}}$

Case B:

$\large\frac{[E_{gene}^{-Ct^{treated}_{gene}}]/[E_{ref}^{-Ct^{treated}_{ref}}]} {[E_{gene}^{-Ct^{control}_{gene}}]/[E_{ref}^{-Ct^{control}_{ref}}]}$

You can rearrange the numerator and denominator to get Case A from Case B. I'm not sure how you are ending up getting different values. Recheck once. It may be because of numerical approximation errors.

Case A looks much neater; so it is best to calculate that way :)