Archiver > GENEALOGY-DNA > 2010-02 > 1266037365

From: "Anatole Klyosov" <>
Subject: Re: [DNA] Variance Assessment wrt back and parallel
Date: Sat, 13 Feb 2010 00:02:45 -0500
References: <>

> From: "Alister John Marsh" <>
> Anatole,
> You seem to be defining/ approaching some things in a different way than I
> have been used to, which may be appropriate to your system, but it may
> have
> been a cause of some of my confusion.

Dear John,

The first part of your sentence - certainly. This is my profession and my
job. I make my living by "approaching some things in a different way" than
other people have been used to. Thank you for noticing it.

The second part of your sentence - certainly, this may be the case. It is

> Brief summaries of my positions are...
> BACK MUTATIONS: My personal experience of finding evidence of back
> mutations
> and parallel mutations moderately frequently in family groups on very fast
> mutating markers gives me a "gut feeling" that back mutations and parallel
> may contribute some affect in the genealogical time frame. Clearly "gut
> feelings" are not recognized as an approved scientific method, but I am
> aware of them never the less.

Let me rephrase what you have said. You have a gut feeling, and you
understand that it is not a scientific category. Good. I am with you.
Now please hear me. You can disregard "back mutations" in family studies.
Their contribution is negligible. This is not a "gut feeling". This is a
statistical "fact".

> GENERATION TIME: I am not sure what you mean when you refer to
> "calibration". Apparently your calibration bears no relationship to data
> from father/ son mutation rate studies.

You are quite right. It is not based on father-son mutation rate studies.
However, it is very close to them. Roughly (sic!) speaking, the father-son
studies give the average mutation rate constant about 0.002 mutations per
marker per generation. My calibrations give 0.00183, 0.00183, 0.00243,
0.00216, 0,00190, 0.0020 mutations per marker per generation (in this
particular case for 12-, 25-, 37-, 67-, 46-, 17-marker haplotypes). My list
contains 31 haplotype formats, and most of them produce around 0.002
mutations per marker per generation for 25-year generation length. Now, can
you really say that "it bears no relationship to data from father/son
mutation rate studies"?

Calibration means that you have a certain genealogical or a historical
event, and you know (or seriously suppose) that you know a timespan to a
common ancestor. For example, you do believe that John Lord of the Isles
lived some 650 years ago. If you set 25 years per generation, 650 years is
26 generations ago. You take all the Donald Clan haplotypes, say, 110 of
them in the 67-marker format, and found that all 110 of them contain 415
mutations from the base haplotype, which is clearly identifiable. Then you
write a simple relationship - 415/110/k = 26, and found that k = 0.145
mutation per 67-marker haplotype, that is 0.145/67 = 0.00216 mutations per
marker. This is a result of the calibration.

Now, you - naturally - want to verify this value. You take a bunch of
67-marker haplotypes of Arabs and Jews, and using this value of 0.00216
mutations per marker as the average mutation rate constant, you find that
the Arabs and the Jews split 4200 years ago. Hmm - you think - this is a
damn close to other independent estimates. Indeed, the patriarch of the Jews
and the Arabs - whatever was his name in reality - lived about 4200 years
ago. So far, so good. Then you checked a number of other historical events
and private genealogies, and it turned out that 0.00216 mutations per marker
per generation is pretty good in description of real events.

Now, tell me, what is wrong with this methodology?

By doing many-many-many examples of this kind I verified the values of
mutation rate constants for 31 haplotype formats. They all fit each other.

Now, tell me, what is wrong with this methodology?

> on this.

I think this is a non-issue at present. With margins of error of some
+/-500 years, who cares about a distance from birth to death of a common

Do you understand your question correctly? Or you mean something else?

>I have a feeling that many forecasts by different people have ignored the
>age of the test subjects, and assumed them all to be 0 years old.

See above. We are not there yet.


Anatole Klyosov

This thread: