Archiver > GENEALOGY-DNA > 2005-06 > 1119300591

From: Robert Stafford <>
Subject: RE: [DNA] was New to DNA - now Statistics of Parallel Mutations
Date: Mon, 20 Jun 2005 13:49:51 -0700 (PDT)
In-Reply-To: <>


There is little need to worry about parallel mutations when determining relatedness for the number of generations we usually consider. Even if one occurred, it would only negate one marker out of 37and bias the call to related. If donors were really unrelated, it should show up on the other markers. There is some effect on TMRCA calculations, but their ranges are so wide that accounting for parallel mutations needlessly complicates the calculation.

Threre is always a small possibility, a few %, that the two people used to determine an ancestral haplotype have coincident mutations in a 10-15 generation genealogy. Again, you would only be off on one marker and it would probably correct itself with later tests.

However, when one marker becomes the sole determinant of a relationship, such as a branch tag, an error becomes 1 out of 1. You can eliminate the possibility of recent mutations by testing other branches. This will probably take care of most cases of parallel mutations. It adds enough confidence to the assessment of an early mutation to make it worthwhile pursuing a connection.

However, it pays to be wary of using common mutations on fast markers to sort lineages without a suporting paper trail. If you find a common mutation on slow markers, say less than a .002 rate, you can probably rely on it. Above .004, I would be a little skeptical, since a number of parallel muttions have been reported for such markers. I suspect that there are a lot more out there that haven't been recognized.

Even if it is shown that there was an early mutation, it doesn't mean everyone who has it belongs in that branch. They should be viewed as pointers to direct further research.

Bob Stafford

Terry Barton <> wrote:
Bob, I am intrigued with your concern about parallel mutations and I am
cautious about my pairs of mutations - in part because of the calculation
that you pointed out.

I am not concerned about the CDYb mutation that identifies a deep split in
our Lineage I - or in the mutation that you don't see on our chart - the 10
at DYS385a. In the 385a = 10 situation, it is 66 to 0 in favor of this
being a basic marker. In the CDYb situation, we have only two results: 37
or 36 - across 25 men tested at that marker. Additionally, where we have
more than one member of a known family tested at CDYb, there is a perfect
correlation to expectations.

I do recall the exercise of checking birthdays and learning that it doesn't
take many people (was it 23?) to find a pair with the same birthday. That
seems to be a correlatation to our having a lot of men in genetic lineage
and fearing that a match at one marker is a parallel mutation. At first
glance, it seems that the two calculations are basically the same. (and
they may well be - but I'd like to hear that from a statistician)

In examining the concern about parallel mutations, how do we consider all of
the other markers that are NOT mutating? Is a 36/37 match any more
conclusive than a 24/25? It seems to me that somehow, the fact that all of
those other markers aren't changing gives more validity to the match being
shared rather than parallel. If I have a 40/42, 41/42, 46/48 or 47/48
match, is it any more likely that the mutation is or isn't a parallel match?
Obviously, the more men tested, the more possibility there is of a parallel
mutation. As we have one of the largest genetic lineages in a surname - we
are at a higher risk.

Is there a point where the number of markers and the closeness of match push
the concern of parallel mutations into a small percentage?

As this concern seems to strike at one of the cornerstones of our usage of
DNA for genealogical purposes, can any of our statistically talented readers
(I guess I shouldn't call them "str"s ) share how we can assess (and
manage) the probability of a shared mutation creeping into a project?

Feel free to use any of my Barton Lineage I numbers as a basis for reference

66 men
34 average markers (approx)
28 mutations (if I have discerned them correctly and there are no parallel
5 generations of unique heritage as an average per participant (this is a
guess - as there are 25 paper trails that don't connect)
4 - number of mutations shared by 2 men (576, CDYa, 444, 438)
2 - number of mutations shared by 3 men (388 is in a known family and 393 is
divided among 2 known families)
1 - number of mutations shared by 5 men (H4 - a known family)
1 - number of mutations shared by 8 men (391 - 5 separate "families")
1 - main branch where 5 men have one result and 20 men have the other - and
no one has anything else (CDYb)
19 - number of mutations which have no match (at least that makes sense -
given other mutations)
1- situation which might be a back mutation (A-37) Or- the 44 mutation may
be a parallel mutation

Here's the webpage where all the results are posted - the earliest known
ancestor is in a left column.

Bob- if your parallel mutation concern should be a guiding element in every
Surname DNA Project analysis - then at what point, if any, can we rule out
parallel mutations? Or is every match without a paper trail totally

I will comment that every match that we have found in the Barton project
"makes sense" from comparing paper trails and locations - except possibly
DYS444 and we are upgrading an existing test to explore that further. (I do
also have one situation in the Hodges Project that seems to be either a
parallel or back mutation.)

Actually, I am as concerned about back mutations as parallel mutations - but
unless the two should be examined together - let's treat them separately.


ps: for the newer readers - a parallel mutation is one that independently
emerged in two different families. It is of particular concern because it
will be a false indication of shared ancestry.

-----Original Message-----
From: Robert Stafford [mailto:]
Sent: Sunday, June 19, 2005 9:38 PM
Subject: RE: [DNA] New to DNA


I raised the issue of branch tags vs. parallel mutations on CDY in the
Barton project in April. You seemed unaware of the statistics of parallel
mutations at the time. This is the thread:

You did not respond to the answers given in the thread, so I am unsure of
your thoughts on the matter. Have you refined your arguments for a
pre-documentation mutation on CDY to take into account the statistics? Have
you eliminated the possibility of a mutation in the documentation period in
each line?

Bob Stafford

Terry Barton wrote:
I have groups of R1b men at both Barton and Hodges who have been tested on
both FTDNA 37 and RG43. I have found more useful mutations in the 5 unique
markers at FTDNA than in the 11 from Sorenson - so those two project's
preference is for the FTDNA37. (I don't want to start a heated discussion -
just wanted to note that all don't share the same opinion about which
testing company's unique markers are more useful.)

Search the US Census Collection. Over 140 million records added in the
last 12 months. Largest online collection in the world. Learn more:

This thread: