PCR errors:How many really?

Tom Chappell dmcbtch at ucl.ac.uk
Wed May 31 12:27:56 EST 1995

In article <3q81hq$h9d at nntp3.u.washington.edu>
yacman at homer26 (L. Bastian) writes:

> Okay, I know papers have been rejected because people have cloned using 
> PCR and then either 1) not sequenced the clones, and/or 2)not tested 
> independently isolated clones for whatever functional assay they are 
> interested in.  The usual line is that using PCR you take the risk of 
> introducing PCR errors into the product.  I seem to recall that Taq 
> polymerase does have a relatively high error rate, about one in ten 
> thousand bases.  For most applications, I think this error rate is really 
> quite low.  Furthermore, anyone who has done any sequencing knows that 
> standard sequencing errors exceed (easily) greater than one in ten 
> thousand.  So, my questions: Anyone care to comment on the 1 in ten 
> thousand rate?  In my, all-be-it, limited PCR experience I have never 
> seen a genuine, confirmed, PCR error.  Am I lucky or what?  Is template 
> selection (e.g. GC content) or amount likely to affect error rates.  My 
> preference is hearing from voices of experience, rather than theoretical 
> gobbeldy gook (sp?)  
> Thanks,
> Scot Bastian Ph.D.

I've seen confirmed PCR errors. There was a construct with 3 PCR
fragments in it that I wanted to be right. I isolated and sequenced 8
individual clones for each fragment. I had one fragment with an error
that was present in 6 of 8 of the isolated clones. The fragment was
only 150 bp long, but it's that error that get introduced in the first
couple of amplification steps that kills you. If I'm PCR'ing a single
fragment for a constuct, I always do 4 independent constructs and make
sure they behave identically. (I'm doing some GFP things right now and
it's fairly easy to tell whether or not the GFP has been messed up in a
way that is critical to the experiment--it's not green...)


More information about the Methods mailing list