quantitative RT-PCR

Srinivas K. Janardan, M.D. vjanarda at umich.edu
Thu Aug 31 12:45:38 EST 1995


In article <421mvm$c6 at miasun.med.miami.edu>, tmiller at newssun.med.miami.edu
(Todd Miller - Pharmacology) wrote:


> This critical assumption, that the amplification efficiencies of the
> target and the competitor are equal, is actually validated (or in most
> cases, *not* validated) by an appropriate plot of the data.  If one
> plots log (Tn/Cn) vs log Co, where Tn is the amount of target after n
> cycles, Cn is the amount of competitor after n cycles and Co is the
> amount of competitor at the beginning of the PCR, one should get a 
> straight line with slope equal to |1| (the sign depends on the choice
> of axes).  Deviations from a slope of |1| show *how* equal the efficiencies
> of amplification of the 2 species were.  If this slope is not |1|,
> quantification is not justified because a fundamental assumption of
> the technique was not substantiated.  Yet if you look in the literature,
> there are lots of plots of this with slopes ranging from 0.6 to well
> over |1|.  I suspect this technique is too new to have been fully
> examined by the scientific community.  The mathematics and more details
> of these considerations can be found in an article by Luc Raeymaekers
> in _Analytical Biochemistry_ 214: 582-585, 1993.  Since the plot is
> made on log-log scales, and the answer is actually the intercept of
> this plot, deviations from the predicted slope = |1| can have huge
> affects on the estimate of To, the amount of target present initially.
> 
> Todd Miller

This statement by Luc Raeymaeykers that the slope must be 1 is correct but
why this is not the case in most publications is not because differences
in efficiences of PCR for the two products.  In fact if you run a single
set of Quantiative RT-PCR reactions and split the final reaction onto two
gels, one agarose and another an Acrylamide gel then calculate the To from
the two gels.  They will be the same in my experience (four different PCR
primers).  Most interestingly the slope on the agarose gels will be
between 0.6 and 0.8 but the slope on the acrylamide gel will be between
0.98 and 1.05.  Thus, I have concluded that the reason for the slope being
less than 1 is due to the agarose gel and not that the amplification
efficiency is different for the two products.  This, I believe, is due to
the high background on agarose gels and no background on acrlyamide gels. 
Why does this matter you ask; light bands for both the standard and the
product of interest are not seen on the agarose gel therefore effecting
the slope but not the point were the two products are equal.

We use this technique to measure what I refer to as "relative"
quantitative RT-PCR.  I do not claim to measure the amount of unknown in a
given sample but rather compare two samples using the same internal
standard, i.e., one sample has x-times the amount of message than the next
sample.

In addition to the slope being 1 there other proofs needed to state that
ones result are valid.  You must get the same results for a given sample
regradless of the amount of starting sample, i.e., you have to demonstrate
that 1x conc.  of your unknown gives the same results as 1/10x conc. (or
more dilute depending on how different conc. all your samples will are)

Next you (may) need to demonstrate that PCR cycle number does not effect
your results.  The experiment needed for this should be obvious.

If you are able to demonstrate this for each PCR primer pair of interest
then I think that you can correctly use "relative" quantitative RT-PCT.

I would welcome any feed-back on the above statements.

I don't take all the credit for this knowledge but got it from a follow-up
letter to Luc Raeymaekers' article written by David Shire entitled
"Quantitative PCR:Reply to L. Raeymaekers' comments" page 59-60 Analytical
Biochemistry (sorry I don't know the volume number.

I also suggest reading "The Impact of the PCR plateau phase on quatitative
PCR" Ciaran Morrison and Frank Gannon, Biochimica et Biophysica Acta 1219
(1994) 493-498.

Now if I could only figure out a way to increase the number of samples I
can get done in a day.



More information about the Methods mailing list