Combining standard error of normalised fold change values - Qbase method - (Sep/08/2015 )

Hi, 

I've used the formula methodology of the qbase software (detailed in Hellemens et al. 2007, open acess) to calculate fold change of my samples using 3 reference genes (geometric mean) and 3 inter-run controls (IRC) for normalisation. In this technique the delta Cq is calculated by the difference of the average Cq of all samples (Cq mean) and the Cq of each sample.

Each step of the calculations combines the associated standard error, for example the standard error from my technical repeats, the error associated with the primer efficiency, the error associated with the reference gene mean (3 refs used) and the error associated with the IRC normalisation. The equations are listed on pages 11-12 of the linked paper. As this is fairly complicated, the software does it all automatically. I'm a broke (no more funds) PhD so attempted this with excel. 

I'm left with normalised values of fold change (relative all samples average) for each biological replicate, each with a standard error. However the description of the method stops there. There is no information on how to work out the stanard error of the average fold change of the biological replicate. 

I'm assuming I do a formula that is similar to the previous accumulation formulas which is (apologies I don't know how to do equations on these post:

Accumulated Standard error of biological replicate mean  =

Squart Root (  ( Standard Error Biological Replicate (SEBR) 1/ Biological Replicate 1)² + ( SEBR 2/ BR 2)² + ( SEBR 3/ BR 3)² + (SE BR Mean/ BR Mean)²  )

Does anyone have any advice on calculating this error of the final normalised biological replicate fold changes?

Cheers 

Rich

I got a reply from Biogazette, the compnay selling the qbase software. Just in case anyone else has this issue I was told to log transform the normalised relative quantities (called CNRQ in the paper) then calculate the arithmetic mean. The SE can then be propagated using the following formula:

Square Root ( Σ SE (biological replicate)2) / number of replicates)

You then back transform both the SE and the mean CNRQ.

Rich   

Hi Rich,

I was looking for the same thing. Thanks for the formula :) Are you considering SD or SE for your statistics?

Also, You can also take the accumulated SD by SQRT((SD1+SD2+SD3)/3