It seems a good time to speak about qPCR
The Polymerase Chain Reaction (PCR) is a method used to synthesize millions of copies of a given DNA sequence.
A typical PCR reaction consists of series of cycles:
This loop is repeated between 25 and 30 times
Why do you want to use qPCR?
(you can replace “mRNA” for other keywords)
Questions are more important than answers
Use the following ingredients
The thermocycler is a computer.
You program it to cook your cake
We simplify and we forget about the polymerase and the dNTP
They both will be represented by primers
The system is represented by this diagram:
The curve depends on the initial DNA concentration
We care only about the exponential phase
The signal increases 2 times on every cycle
\[X(C) = X(0)⋅2^C\]
So we can find the initial concentration
\[X(0) = X(C)⋅2^{-C}\]
DNA concentration crosses 50% at 13.73 cycles
Start with a large concentration of template, and dilute it several times. Measure the CT of each dilution
If everything is right, we get
\[X(0) = X(CT)⋅2^{-CT}\]
But sometimes we do not know \(X(CT)\)
because the Signal is not 100% DNA concentration
It depends on the fluorophore assimilation
We still can measure change of concentration
Let’s say we extract mRNA before and after a shock
\[\begin{aligned} X_B(0) & = X_B(CT_B)⋅2^{-CT_B}\\ X_A(0) &= X_A(CT_A)⋅2^{-CT_A}\end{aligned}\]
therefore the fold change of expression is
\[\frac{X_B(0)}{X_A(0)} = \frac{X_B(CT_B)⋅2^{-CT_B}}{X_A(CT_A)⋅2^{-CT_A}}\]
If we assume that the DNA concentrations are the same when the signal crosses the threshold, i.e.
\[X_B(CT_B)=X_A(CT_A)\] then \[\frac{X_B(0)}{X_A(0)} = 2^{-(CT_B-CT_A)} = 2^{-Δ CT}\]
Here \(Δ CT\) means the change in CT for one gene in two conditions
This change of concentration has two components
To avoid the second component, we use an endogenous reference
(typically, a housekeeping gene)
A single pipet, at the same time
We normalize each sample
\[\begin{aligned} \frac{X_B(0)}{R_B(0)} &= \frac{X_B(CT_{XB})⋅2^{-CT_{XB}}}{R_B(CT_{RB})⋅2^{-CT_{RB}}}= K_B⋅ 2^{-(CT_{XB}-CT_{RB})}\\ \frac{X_A(0)}{R_A(0)} &= \frac{X_A(CT_{XA})⋅2^{-CT_{XA}}}{R_A(CT_{RA})⋅2^{-CT_{RA}}}= K_A⋅ 2^{-(CT_{XA}-CT_{RA})} \end{aligned}\]
\(K_A\) and \(K_B\) are constants that depend on the target and reference genes, and how each Signal changes with concentration
\[\frac{X_B(0)}{R_B(0)}÷\frac{X_A(0)}{R_A(0)} = \frac{K_B}{K_A} ⋅ 2^{-(\Delta CT_B-Δ CT_A)}\]
We can assume that \(K_B=K_A,\) because we are comparing the same pair of genes every time
In that case the change in relative expression is
\[\frac{X_B(0)}{R_B(0)}÷\frac{X_A(0)}{R_A(0)} = 2^{-(\Delta CT_B-Δ CT_A)}\]
It is usually a good idea to take logarithms
Using \(\log_2\) we get log fold change, which can be written as
\[\begin{aligned} Δ CT_A - Δ CT_B &= \overbrace{(CT_{BX}-CT_{BR})}^{\text{before}} - \overbrace{(CT_{AX}-CT_{AR})}^{\text{after}}\\ &=\underbrace{(CT_{BX}-CT_{AX})}_{\text{target}} - \underbrace{(CT_{BR}-CT_{AR})}_{\text{reference}} \end{aligned}\]
This works very well with a linear model
The reordering of deltas is equivalent to \[\frac{X_B(0)}{R_B(0)}÷\frac{X_A(0)}{R_A(0)} = \frac{X_B(0)}{R_B(0)}⋅ \frac{R_A(0)}{X_A(0)} = \frac{X_B(0)}{X_A(0)}÷\frac{R_B(0)}{R_A(0)}\]
In other words, the ratio of normalized values is also the ratio between ratios of change
(I think the formula is more clear than the text)
In that case the standard curve slope changes
If the primer efficiency is \(E_X,\) the correct formula is
\[\frac{X_B(0)}{X_A(0)} = \frac{X_B(CT_{BX})⋅(1+E_X)^{-CT_{BX}}}{X_A(CT_{AX})⋅(1+E_X)^{-CT_{AX}}} =(1+E_X)^{-(CT_{BX}-CT_{AX})}\]
The efficiencies may be different for the housekeeping gene
\[\frac{R_B(0)}{R_A(0)} = \frac{R_B(CT_{BR})⋅(1+E_R)^{-CT_{BR}}}{R_A(CT_{AR})⋅(1+E_R)^{-CT_{AR}}} =(1+E_R)^{-(CT_{BR}-CT_{AR})}\]
\[\frac{X_B(0)}{X_A(0)}÷\frac{R_B(0)}{R_A(0)}= \frac{(1+E_X)^{-(CT_{BX}-CT_{AX})}}{(1+E_R)^{-(CT_{BR}-CT_{AR})}}\]
The log ratio is \[-F_X(CT_{BX}-CT_{AX})+ F_R(CT_{BR}-CT_{AR})\] where \(F_X=-\log_2 (1+E_X)\) and \(F_R=-\log_2 (1+E_R)\) are the slopes of the corresponding standard curves
We know the dilutions, the CT values, and their relationship \[X(0) = X(C)⋅(1+E)^{-C}\] That is \[\text{initial}_i = \text{threshold}⋅(1+E)^{-CT_i}\] Taking logarithms, we have \[\log(\text{initial}_i) = \log(\text{threshold}) + \log(1+E)⋅ -CT_i\] Notice that threshold concentration should be the same for all \(i\)
We can find \(E\) using a linear model like \[\log_2(\text{initial}_i) = β_0 + β_1 ⋅ CT_i + e_i\]
Fitting the linear models gives us \[β_1 = -log(1+E)\] and from there we can find the primers efficiency \(E\)
For the first set of primers, we have
2.5 % 97.5 %
(Intercept) 0.288 0.332
CT -0.694 -0.692
We get intervals for \(\beta_0\) and \(\beta_1.\) The efficiency is
2.5 % 97.5 %
1.000 0.997
In this case the real value is 100%
For the second set of primers, we have
2.5 % 97.5 %
(Intercept) 0.108 0.566
CT -0.606 -0.582
The efficiency is in the interval
2.5 % 97.5 %
0.833 0.790
The real value is 80%
It is recommended to always do the standard curve by triplicate
All our measuring devices have a margin of error.
There may be a small error measuring the signal.
That will affect the resulting CT value
When the initial concentration is small, the signal may not reach 50% in 30 or 40 cycles
Then we will use a lower threshold, and pay the price
Since the curve is nearly flat in the lower signals, a small error in the signal has a large impact in the concentration
Thr | CT | Concentration | Error.ratio |
---|---|---|---|
0.05 | 10.00 | 1025.1 | |
0.06 | 10.21 | 1186.4 | 0.1573 |
0.50 | 13.73 | 13624.5 | |
0.51 | 13.78 | 14054.3 | 0.0315 |
Thus, an error between signal 0.05 and 0.06 results in an apparent increase of 16% in concentration. But an error between signal 0.5 and 0.51 results in an error of 3%
Livak KJ, Schmittgen TD. “Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method”. Methods. 2001 Dec; 25(4):402-8. doi: 10.1006/meth.2001.1262
Pfaffl, Michael W. “Relative Quantification.” In Real-Time PCR, Published by International University Line (Editor: T. Dorak), 64–82, 2007.