Archiver > GENEALOGY-DNA > 2010-02 > 1266294148

From: "Alister John Marsh" <>
Subject: Re: [DNA] TMRCA assessments
Date: Tue, 16 Feb 2010 17:22:28 +1300
References: <><05039ABA2BCE42B8B72DB004F08F05D6@anatoldesktop>
In-Reply-To: <05039ABA2BCE42B8B72DB004F08F05D6@anatoldesktop>


You said...
David mistakenly assumed (using a "simple maths") that 50% of all
mutations should be back mutations (at least I understood him that way), but

it is incorrect. In the same vein, John suggested that in his series he saw,

or "assumed", or "believed" a significant contribution of back mutations,
however, there were no many double mutations in his dataset. Actually, there

were two, as I remember, but they belonged to a different branch. Are they
considered as "parallel mutations", I wonder??

If a marker mutates once, and has a second mutation after a period of time,
the second mutation has roughly 50% chance of being a back or forward
mutation. Perhaps that was what David was meaning. Is that not correct?

In the case of what I said, I suspected there was 1 "back mutation" in the
dataset I gave a few days ago. I am awaiting further DNA marker results
which may increase the evidence that that was indeed a back mutation. On
what I can see in that particular dataset, the suspected back mutation may
have changed the mutation count by 1, or 2 mutations (using your method),
depending on how the ancestral haplotype is assessed.

The back mutation may therefore have changed the mutation count in that
example by the order of 10%, give or take a bit. If I was asked to
mischievously place that back mutation on the tree at the place where it
could cause the maximum amount of havoc/ confusion, I could no doubt place
it where it would cause a significant impact on mutation count. Given your
assurance that statistically back mutations don't normally occur in the
genealogical time frame, it is interesting that in the one example I put
before you, there was likely 1 back mutation, and perhaps more if we could
only just see evidence of them from the context.

I suggest that you have to think twice about potential errors of 10%,
because if you have other errors of 10% or 20% or 100% compounding with
them, they become part of a troublingly large confidence interval.

What I have been saying is that in the example dataset I provided, the
impact of alternative ways of interpreting ambiguities created by the
"parallel mutations" has had a much greater impact on the mutation count
then the back mutation had. The problem being that the back mutations may
have contributed to incorrect identification of the ancestral haplotype. If
the ancestral haplotype is misidentified (or at least made
ambiguous)"because" of parallel mutations, then contrary to you claims, they
are not just the same as any other mutation, but mutations which
specifically can lead to large changes or ambiguities in the mutation count.

"Ordinary mutations", if I can call them that, ie mutations which are not
back mutations or parallel mutations, do not normally create the same
opportunities for misidentifying the ancestral haplotype. So that is why I
believe back mutations can't be just brushed off as merely "ordinary
mutations" to be treated no differently than any other.

In a recent posting I said...
" As I understand it, back mutations could mess things up in Anatole's
method if they were at inconvenient places, but Anatole says they are rare
in the genealogical time frame. In most cases they are much rarer than
parallel mutations in the genealogical time frame, so he is probably right
that they are less significant in the majority of genealogical time frame


This thread: