I have a question regarding to the representation of the relative expression histograms. We did a comparative qPCR study, and, using the Pfaffl's efficiency corrected Ct formula, I calculated the relative expression values for all of my samples. Since I have biological replicates, I get error bars for the treatment groups on the relative expression values, but in many papers I noticed that they also use error bars for the control groups (value 1 with an error bar). How?
Biogazelle's qbasePLUS software (http://www.qbaseplus.com) can do this. If you want to do this manually in a spreadsheet, I would need a bit more information on how EXACTLY you did you calculations. If you calculated relative quantities for all you samples at once (e.g. according to Hellemans et al., Genome Biology, 2007), you will also have variable results for your control group and hence an error bar for this group.
Dr. V, your software is the best in the world for this. It's good to let the cat out of the bag at this point. And it is always good to hear your opinion on qPCR stats. The error bars for the controls is always propotional to their error bars before the control was divided by itself. I believe this is also equal to the coefficient of variance. E.g., if the error bar for a control value of 0.5 was +/- 0.2 (before control was divided by itself), then, when the control becomes "1" by self-division, the error bar becomes (by proportion or coefficient of variance rules) +/- 0.4. But if the error bars are the result of transformation from log to linear scale, the error bar above and below the median is not symmetrical... technically, and thus a more lengthy explanation is needed. Error just doesn't simply disappear... must always be accounted for - and very tricky, unless you use Dr. V's and Dr. H's software.
I do not see the point of showing the "normalized controls" as a bar of height 1 - with or without error bars. This value (rel. conc = 1) should better be represented a a horizontal line. Actually, it is the x-xasis to what the other values refer to. This is what thie normalization does.
Form the "treated" group(s), you get a delta-ct for each sample, similarily for the "control" group. Then you average the delta-cts for each group. These averages have errors (can be expressed as IQR, MAD, SD, SEM or what you like). Now to calculate the (average) difference between a treated and the control group, you calculate the difference between the delta-cts, what gives the delta-delta ct. Since just two averages are subtracted here, it is not straight-forward to see how the error is to be calculated. The correct way is to use error propagation for SD or SEM. This error is an error of the differences. Now showing the differences together with one of the values that has been used to get this difference is not quite sensible.
But if this is done that way, the error of the "control" group (with average 1) is calculated the same way (error propagation) with twice using the error of the delta ct of the control group.
- Badges
- Science topic