GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2007-05 > 1178112335


From:
Subject: Re: [DNA] Double or single cousin?
Date: Wed, 2 May 2007 09:25:35 EDT


In a message dated 5/2/2007 2:11:00 AM Pacific Daylight Time,
writes (to Kathy):

> Except that you've missed what I meant.
> You have 1/2 your mother's genes. But anywhere between 0 and 1/2 your
> grandmother's genes. Whilst you would expect to have about 1/4 of her genes
> that is not certain - and adding up all those grandparents' genes is not as
> simple as you suppose - even though the one thing you *do* know before you
> start is the answer!:
>
> grandparentA + grandparentB + grandparentC + grandparentD =1
> A + B = 1/2
> C + D = 1/2
> But you don't know A, or B, or C, or D.


The whole thread is about "expected" values, with the implicit understanding
that we don't always get what we "expect" (exactly like coin tosses). But I
think raising the specter of getting 0 or 50% of a grandmother's genes might be
more misleading to the casual reader than simply stating the expected or
average value. John Chandler's recent message was more useful in that regard,
although I had set it aside to ponder one of the parameters.

=== begin quote from John Chandler

Because each chromosome is inherited separately, and each inherited
chromosome is a patchwork of pieces from the two chromosomes that
the parent has, we can subdivide the comparison between two siblings
into something like 5 x 46 separate comparisons of independently
inherited chunks of DNA.  If we agree to call the four parental
copies of each chromosome "completely different" for the purposes
of assessing heredity, then the probability distribution of the
two siblings' level of matching will be closely approximated by
a gaussian with a mean of 50% and a standard deviation of 3%.

In other words, with 95% confidence, you can expect any two siblings
to share between 44% and 56% of their DNA.  In principle, they could
share 0% or 100%, but that would be highly unlikely.

===== end quote from John Chandler

Using five for the number of chunks struck me as a bit high, but I haven't
yet had time to find citations about the average number of cross-over points in
recombination, so I'm not offering a substitute. In any event, it would just
change the confidence limits slightly.

Now, John Cartmell (I can't use the abbreviation John C any more!), I don't
for a minute believe that Kathy was supposing the process was "simple", as you
charge -- but do you have some suggestions of your own on how to model it in a
more realistic way?

Ann Turner


**************************************
See what's free at
http://www.aol.com.


This thread: