GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2010-02 > 1266209667


From: David Ewing <>
Subject: Re: [DNA] Variance Assessment wrt back and parallel mutations
Date: Sun, 14 Feb 2010 21:54:27 -0700


Anatole, I understand very well that to work out actual placement of
haplotypes in a family tree requires conventional genealogy. In attempting
to do that, I discovered that parallel mutations were surprisingly common.
You told me that they were so rare as to be negligible in the time frame
that surname projects are concerned about. I asked you to show me.

You provided a tree. One third of the mutations in the first and second
branches you found were parallel mutations. I showed you this explicitly by
labeling part of the diagram you sent:
http://dl.dropbox.com/u/431003/EwingTreeByAnatole-2.pdf

Now, I have worked out which mutations your phylogeny program used to
delineate the third branch you found (which corresponds to my Group 2). You
can see this at:
http://dl.dropbox.com/u/431003/EwingTreeByAnatole-2.pdf
Your program adduces ten parallel mutations among the twenty seven it uses
to define sub-branches in this tree.

I told you parallel mutations were common; you argued that they were not. So
far, you have shown me that 1/3 of the mutations in our tree are parallel
mutations. This is a smaller percentage than the roughly 1/2 of all
mutations that I was finding as I made trees informed by conventional
genealogy, but it does not appear to me to be "negligible."

This discussion is not about exact placement of haplotypes in a family
tree--I know you cannot help me with that. It is about how to explain that
there are so many parallel mutations. Your first response told me, "Good
news," and suggested that you had solved the problem. I have knocked my head
against this problem enough that I can now assert that a tree accounting for
the Ewing data cannot be drawn without including a relatively large number
of parallel mutations, even if this is done without regard to known
conventional genealogic connections.

You do not like imprecise language like "relatively large." Let me be
precise. You cannot account for the Ewing data without adducing at least 25%
of all mutations as parallel mutations. I should be very happy to be proved
wrong, but this cannot be done by showing me a tree with 33% parallel
mutations.

David Ewing


This thread: