Archiver > GENEALOGY-DNA > 2012-02 > 1329363999

From: "Lawrence Mayka" <>
Subject: Re: [DNA] Out of Africa
Date: Wed, 15 Feb 2012 21:46:43 -0600
References: <><>
In-Reply-To: <>

1) I'm glad we both agree that modern DNA test results alone cannot determine with certainty where our Y-Adam lived. Other evidence must be taken into account as well.

2) Calculations using the spreadsheet's full set of 55 markers (it already skips some particularly problematic ones) will indeed underestimate long intervals to some degree. Your own explanation was quite garbled, but the better explanation is that each STR apparently has a natural lower and upper length limit--or at least some centering tendency--and hence over a long period of time is restrained from performing a true random walk (as most stepwise numerical models assume). The net effect is to artificially reduce the mutation count difference between compared haplotypes; and this effect is more pronounced on a faster marker, which is likely to hit a natural limit sooner. The tradeoff is that the confidence interval increases (i.e., the statistical accuracy of our result decreases).

With the full set of 55 markers, the interclade TMRCA between A1a and A3a was 75,000 +- 16,000 years.

I dropped an additional 7 markers whose published mutation rates exceed 0.005/generation. For the same A1a/A3a TMRCA, the remaining 48 markers calculate 88,000 +- 20,000 years.

I then dropped an additional 10 markers whose published mutation rates exceed 0.0025/generation. The remaining 38 markers calculate a TMRCA of 115,000 +- 29,000 years.

Next, I dropped an additional 5 markers whose published mutation rates exceed 0.002/generation. The remaining 33 markers calculate a TMRCA of 130,000 +- 35,000 years.

Finally, I dropped 7 more markers with published mutation rates exceeding 0.001/generation. The remaining 26 markers calculate a TMRCA of 160,000 +- 49,000 years.

Keep in mind that the confidence intervals (plus-or-minus values) assume that the mutation rates themselves are perfectly known. In actuality, we can state as a general rule that the lower the mutation rate, the more uncertain it is--simply because the measured rate of an event so rarely observed has itself a large confidence interval.

There is another inaccuracy which becomes more prominent at lower mutation rates: the assumption that all mutations are single-step. Any multistep mutation artificially inflates the calculated TMRCA, and this effect is more pronounced on markers with lower mutation rates.

The bottom line, frankly, is that STRs do not make a very good clock for such long periods. We need to look toward the use of SNPs for long-period clocking.

> From: [mailto:genealogy-dna-
> ] On Behalf Of
> A1. It seems that you now agree that we do not know that the common ancestor of
> Europeans lived in Africa. If so, this is correct.
> Frankly, I expected your comment on your grossly incorrect "calculations" using 67
> marker haplotypes (or whatever available) and obtaining 70,000 ybp. Do you
> understand now that it was wrong? You can simply say "yes, now I understand", or
> "no, I do not understand".

This thread: