PCR fidelity: Calculations?

Brian Foley brianf at med.uvm.edu
Mon Oct 24 18:03:50 EST 1994

	In Nucleic Acids Research 19, 6052 (1991) Yuhong Zhou, Xiaoping 
Zhang and Richard H. Ebright say they found 35% of their 633 bp PCR
products had a PCR misincorporation.  They then state the error rate as 
3.7 x 10^-5.
	In Nucleic Acids Research 18, 3739-3744 (1990) Kristin A. Eckert and 
Thomas A. Kunkel state that the arror rate can vary from 2 x 10^-4 to 
less than 1.2 x 10^-5 mutations per base per cycle.
	In the Journal of Biological Chemistry 264, 7780-7783 (1989) 
Supan Fucharoen et al, sequenced five clones of a 3,000 bp PCR product 
and found no misincorporations.  They state that "In our experience, 
there can be misincorporation when the smaller (e.g. 630  bp) amplified 
fragment has been cloned and sequence-analyzed (data not shown).  In this 
paper when as large as the 3.0 kb amplified fragment was cloned, complete 
sequencing of five independent clones (14,990 bp) derived from normal DNA 
revealed no error, therby suggesting that the amplified products of the 
relatively large size are less likely to be misincorporated during DNA 
synthesis with Taq DNA polymerase."

	I don't beleive that larger PCR products are be less prone to 
errors than short ones.  I believe that the error rate for PCR is quite 

	The question I have is:  How do we calculate the "error rate per
base per cycle"?
	Zhou et al found 7 point mutations in 20 clones of 633 bp after
30 cycles.  That is 12,660 bp sequenced and 7 mutations found.  They say 
that translates to "an error rate of 3.7 x 10^-5 per base (formula in ref 
3)".  I do not have access to ref 3 which is:  Eckert,K. and Kunkel,T. 
(1991) in Quirke,T. and Taylor,G. (eds) _Polymerase Chain Reaction I:
A practical Approach_, IRL Press, Oxford, U.K. in press.
	I assume that the formula is quite complicated.  It would seem 
from my experience that it would depend to some degree upon the amount of 
template supplied.  But in theory I think it should not.

	None of the references above used methods such as high dNTP 
concentration or high mg++ concentration to increase the error rate.  So 
it seems that the rate can vary even under "standard" conditions.  On 
February 21, 1994 Eric First (erfi at eel.sunet.se) posted a summary to this 
group "Summary: Fidelity of Thermostable DNA Polmerases" which had many
references for Taq fidelity:

	1.1 x 10^-4   Gene 112, 29-35 (1992)
        2.1 x 10^-4   Proc. Nat. Acad. Sci. 86, 9253-57 (1989)
        8.9 x 10^-5   Nucl. Acid. Res. 15, 4193-98 (1991)
        2.0 x 10^-5   Gene 108, 1-6 (1991)

and many more...

	This is slighly more than a 10-fold spread.  But the implications
for success or failure of a project when 35% of a 633 bp product have 
error(s) make you want to look into this further...

*  Brian Foley               *     If we knew what we were doing   *
*  Molecular Genetics Dept.  *     it wouldn't be called research  *
*  University of Vermont     *                                     *

More information about the Methods mailing list