Please explain how the p-value and SD is calculated using Cp values.
Dear Ranjuna,
Your question is unclear. In my experience, p-value and SD are related to the statistical analysis. If you want to analyze the Cp, extract the value of Cp, put in an excel file and then you can do the analysis of p-value by statistical software. Good luck.
Ok, the most important thing here is not to calculate the SD of Cp values.
If you're calculating it between replicate PCRs, then all you're doing is semi-quantifying your pipetting error. You should not do this. If pipetting error is a relevant concern, don't measure it, just...pipette better.
If you're calculating it between the average Cp values of biological replicates, then this is still inappropriate because you're using a linear measure to assess an exponential variation. Also, you won't have normalised your data by this point, either.
So (and bear with me here, because this is fairly laborious):
You need to convert your data to linear, normalised values before you do any statistical analysis on it (I favour linearising everything separately and then normalising, but using a delta-delta Ct method and then linearising should also work).
I would thus do the following: take your mean Cp values (i.e. the average of the three sample replicates, assuming you're doing all PCRs in triplicate, and you have acceptably low pipetting errors) for each of your biological replicates, and then linearise these to derive relative levels using the following equation:
expression level of sample = efficiency(lowest Cp - sample Cp)
Efficiency of your reaction here is slightly counter-intuitive: it's effectively the "factor by which your reaction product increases each cycle", so for a 100% efficient reaction the amount of product doubles each time, thus the value you use is 2. Your PCR should be around 95-100% (you can check this via a dilution series -Cp values should drop by 3.32 every ten-fold dilution of sample), so we'll assume efficiency here is 2.
So if you have three biological replicate samples (A, B, C) with Cp values (for your gene of interest) of 24, 26 and 25, and efficiency is 2, you have
2(24 - 24) = 1
2(24 - 26) = 0.25
2(24 - 25) = 0.5
So you now know that A has the most transcript, C has half as much as that, and B has a quarter.
Now do the exact same with your Cp values from your reference gene*, which we shall say result in values of 1, 0.3 and 0.6 for A, B and C respectively.
So then simply do GOI value A / Reference A, GOI value B/Reference B, and so on, thus 1/1, 0.25/0.3, and 0.5/0.6 gives you 1, 0.83 and 0.83.
THESE values can then be used to obtain a standard deviation, as they reflect the linear variation in expression level of that gene in three independent biological replicates.
So if you have test samples (3 biological replicates) and control samples (3 biological replicates) and you do the above for all samples, you might end up with 1, 0.83, 0.83 for your control, and 0.05, 0.1 and 0.12 for your test sample, so you can average the three controls, vs the three tests, or use the two sets in a statistical test, or whatever.
This doesn't, of course, tell you how many transcripts you have, because everything is simply expressed relative to the highest observed expression level, but most qPCR data is reported as fold changes anyway, so that's all you need.
*I would actually use multiple reference genes, and I would establish which genes to use via screening a fairly large panel of candidates. You then generate a linear normalisation factor by calculating the geometric mean of your linear values for each reference gene (i.e. not "(Gene1+Gene2)/2" but "square root of gene1*gene2").
Hi John, Thank you very much for your detailed answer. Very helpful.
Do you mind explaining a bit about how the p-value is calculated? Which numbers to use and how?
- Badges
- Science topic