GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2010-03 > 1269793301


From: David Faux <>
Subject: Re: [DNA] 23andme cousin results
Date: Sun, 28 Mar 2010 09:21:41 -0700
References: <CCC182DA-E426-430A-BA37-355BD4D5E8AD@earthlink.net><ea3bd9561003272128k3b567d83y190aeb937b77803b@mail.gmail.com><253CFBB8-15B8-4B84-81E5-A4F86DEEC8D7@earthlink.net>
In-Reply-To: <253CFBB8-15B8-4B84-81E5-A4F86DEEC8D7@earthlink.net>


John,

The two charts provided by FTDNA on the FAQ page for their Family Finder are
extremely helpful in understanding complexities such as how much a thrid
cousin twice removed "should" share with you. They are found here:
http://www.familytreedna.com/faq/answers/17.aspx

I am still doing my calculations as to "value added" scenarios with respect
to the complex family constellations I have in my project. Clearly to
the third cousin level you are likely to get pretty close to the "expected
sharing". A third cousin and I (we share the same great great grandparents
Sgt. Charles Warren Clarke Faux and Mary Ann Williams) are expected to share
0.78% and our observed value is 0.83%. I am finding that with my much more
mixed Young family, the results are somewhat dodgy with a number of outliers
(one in particular). By the time that the fouth cousin level is reached it
is not unusual to share nothing - hence the need to find as many family
participants as possible so that by chance some will have some degree of
sharing - even quite significant especially in my Young study. It is all
breaking new ground as far as I can see. As far as I know there is no
academic study to which we can refer for answers.

My results will be ready for public display in a few days and it will have
expected versus observed values with an attempt to note when there is a
cousin relationship in the picture. In my work the added "cousin factor"
would likely be fairly small since the participants with known cousin
connections are no closer than second cousins to each other, and the cousin
marriages are at least two generations beyond either. I am assuming that an
uncle and nephew (we have two in the study), for example, would share the
same amount whether their great or great great grandparents were first
cousins plus or not. Based on limited observations this would seem to be
correct.

In my case the cousin marriages (two) are not found until the great great
grandparent level. In terms of those who are related, I am expected to
receive 1.56% from each of my 64 ggg grandparents but 3 are cousins so 4.68%
of my autosomal structure should have come from the "Young family" - but
each of these would have received varying percentages from the family
partriarch Adam Young and his wife Catharine Elizabeth Schremling so it
gives me an unknown boost in that line. When I reach the generation which
includes this couple directly (e.g., 5th great grandparent) I can work out
the expected percentages but now we are getting into the "drop off"
territory where some ancestors in the family tree are genealogical ancestors
but not genetic ancestors. I can see why some people shy away from this
sort of work.

One factor which I have not seen addressed though is the reported female to
male ratio of 1.6 to 1 in terms of recombination rate. Thus I would expect
to see greater sharing between those in the direct male line due to a
reduction in the number of recombination events. I don't recall this being
discussed in the present context. All figures such as expected sharing from
a great great great grandparent are given as a single percentage figure
whereas, if I understand the process correctly, there "should" have more
sharing with those in the direct male to male line than cousins at the same
(e.g., thrid) level via the female lineage - but why has no one spoken of
this - perhaps I am missing something fundamental (would not be the first
time).

David K. Faux.

On Sun, Mar 28, 2010 at 12:15 AM, John Carr <>wrote:

> David, Tim, others as well,
>
> Are the genome wide SNP shared segments within known cousin relationships a
> fraction of a percent as 23andme shows for our so called cousin matches or
> are the shared segments usually larger?
>
> Given that the whole genome SNP results for two siblings from the same
> mother and father could range between having all segments match to no
> matching segments, with the greater frequency of matches within sibling
> relationships being closer to 50% than 0% or 100%, it would seem that there
> would be a wide variation of matching across the genomes of biological 1st
> to 5th cousins. While 100% matching is less likely, the occurrence of 0%
> segment matches should have a higher occurrence rate than for siblings.
> The relatedness of the DNA donating ancestors would surely affect the
> result, a succession of first cousins being more likely to share closer to
> 100% than when the DNA donors shared ancestors are several generations
> removed. Your studies look at situations where there are a lot of first
> and second cousin parents, thus the expectation is that you will find a high
> percentage matching genome frequencies within the group. I would expect
> that actual 4th to 5th cousin rela!
> tionships for most people whose ancestors were not first cousins would
> still have a higher occurrence rate than a fraction of a percent. Though, as
> I stated, there will be some with near 0% matching segments.
>
> So, the question is, what do practical studies show? Are there sufficient
> studies of three or four generations within a family to determine the modal
> frequency with which SNP segments can be expected to be shared between the
> grandparents and the grandchildren?
>
> How does this vary between situations where the parents are close cousins
> and situations where the parental ancestries are separated by thousands to
> tens of thousands of years?
>
> John
>
>


This thread: