So here is what I find reasonable to calculate the standard error: You calculate the propagated standard deviation s^2= s1^2 + s2^2, in which s1 and s2 are the standard deviation of housekeeping gene and gene of interest Ct values. Then you get standard error: SE = SD/sqrt (number of replicates). So I have two questions:
- If you use 2 or more housekeeping genes, how will the standard deviation be calculated now? Would you add the variance of the other housekeeping gene to the sum above or get the variance of the geometric values of 2 housekeeping genes?
- What is the number of samples if you lose value of either your housekeeping genes or GOI? Say I have 2 values for GOI and 3 values for HK?
Thanks!
The measurements of the reference gene and the gene of interest are correlated ("paired"). If they were not, then it would not make sense to measure the reference gene at all! The reference gene serves as a "loading control" and it is used correct the readings for the gene of interest between different samples for differences in sample preparation (sample conc., RT efficiency, etc).
Because of this correlation, the standard deviations or -errors of the Ct-values are not very interesting. What is interesting is the variability of the delta-ct values (the differences between the reference gene ct and the ct of the gene of interest within a given sample). These delta-ct values should not be correlated anymore.
If you have several reference genes you can simply use the average of the ct-values of all these genes as your loading-control proxy. This requires that all reference genes were measured in all samples. When this is not the case you may irgnore the reference gene(s) that were nor measured in all the samples. If only few values are missing, an imputation of the missing values could help so that fewer data is wasted. Another possible option (I have not carefully thought through it!) might be to use a mixed model and model the reference gene expressions as random effects (crossed with sample-ID)...
So because I have run several plates and my housekeeping genes were tightly close so I was told not to run them again, just save the templates to run just the gene of interests. And for one GOI, the ct value was high so I got 2 values of 38 and one No ct. That explains why I have 3 replicates of Ct value for HK but 2 for GOI. I thought you can calculate the average dCt = average of Ct(GOI) - average of Ct (HK)?
No, I don't see how this would make any sense.
If your HKs were so close together that you decided not to measure them for all your samples, then the only conclusion is that you consider your sample preparation as sufficientily standardized and reliable so that you can rely on equal amounts of input material and that you thus do not need the correction for a loading control.
This means you can forget the measurements for the HKs all together!
- Badges
- Science topic