GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2004-03 > 1080766104


From:
Subject: Re: [DNA] Some thoughts about the near future
Date: Wed, 31 Mar 2004 15:48:24 -0500 (EST)
References: <20040331192417.89747.qmail@web20005.mail.yahoo.com>
In-Reply-To: <20040331192417.89747.qmail@web20005.mail.yahoo.com>(peterebay@yahoo.com)


Peter wrote:
> Is it possible to estimate the percentage of close
> matches that are expected to be due to convergence and
> the percentage of close matches that are expected to
> be due to relatedness (ancestry by descent)?

Perhaps, but first you need to define exactly what you mean by
a close match and relatedness. For example, you said later
that a close match is 22/25 or better, but you didn't say
anything about what constitutes a relative and what doesn't.
If you accept the notion that all existing human Y chromosomes
descend from a "Y Adam" who lived on the order of 100,000
years ago, then we are all related, no matter what. If two
people are "similar by descent", does that mean their
respective contemporary ancestors have always been at least
that close a match, back to the common ancestor? Or does it
mean neither line has ever had a mutation cancel out, and
neither line has ever had a mutation that happens to parallel
a mutation in the other line? Or does it mean simply that the
common ancestor was no more than X generations ago?

> we be so sure that the various clusters of near
> matches are not mostly due to Ghengis Khan type
> situations or the prolificness of the those lines?

There is a way to tell the difference -- make a network
diagram of all the haplotypes in the target population.
If they form a dense grid, then the common ancestor is
too far back in time to matter. If they form a "star",
then the cluster could indeed be a descent group with an
identifiable, recent common ancestor.

> Regarding my statment that it is predicted that at
> about 140 STR markers, a y-chromosome test becomes a
> personal identity test, what I was told was "Using
> statistical models, the TMRCA approaches 1 as the
> number of markers reaches about 140 and could
> therefore work as an identity test at that level."

That's just plain silly. An identity test has to have
a TMRCA approaching 0, not 1. What's more, the median
is not good enough for an identity test -- for forensic
purposes, a test would have to show odds of at least a
billion-to-one against false positives to qualify as an
identity test.

John Chandler

This thread: