GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2002-08 > 1029449580


From: "John F. Chandler" <>
Subject: Re: [DNA] Test Variability
Date: Thu, 15 Aug 2002 18:13 EDT
In-Reply-To: huffaker@chartertn.net message <002101c23997$9e7ad450$19799e18@huffaker> of Thu, 1 Aug 2002 14:22:39 -0600


Jim wrote:
> My brother and I were one count different at Y-GATA-A10. Thirteen for me, 14
> for him. All the other loci counts were identical. The testing company just
> remarked that it was unusual for brothers to not have identical counts at
> all loci. My background is Analytical Chemistry and had my laboratory
> produced "unusual" results, we would likely re-do the test over and again.
> (And maybe they did).

Yes, they probably did. It seems likely that they made sure their
"unusual" result was reproducible.

> Getting to the point here, what is the reproducibility of the test method(s)
> from which counts are determined?

The paper by Wilson et al. in the Proc. Natl Acad Sci reported a test of
reproducibility. They found 0.7% discrepancies between two independent
series of tests of the same subjects. That would imply an error rate
of roughly 0.35% on average. It's hard to say whether that's typical
of the commercial labs or not. We do hear occasional reports of results
being corrected, but I don't think anyone is keeping score.

> What is the 95% confidence interval of the range? These seem important
> questions to me because the test results also conclude a common ancestor for
> another participant and I and he and I had one count difference at DYS 390
> and two count difference at DYS 392, from which it was concluded our common
> ancetor lived in 1690 about.

Huh?? That's entirely too precise. It's also overoptimistically recent,
but that's probably due to their "in-house, unpublished" estimate of the
mutation rate that's twice as high as the bench-mark everyone else uses,
plus a swashbuckling assumption that a two-step difference can be treated
the same as a one-step difference. You need to recognize that the error
bars on that figure are multiple centuries.

John Chandler

This thread: