GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2007-03 > 1173917193


From: "Jim Pearsall" <>
Subject: Re: [DNA] OmniPop caution received from FamilyTreeDNA
Date: Wed, 14 Mar 2007 19:06:33 -0500


----Original Message Follows----
From: (John Chandler)

Al wrote:
> Where is the fault with the method?

We discussed this a bit last time. The numbers that come out of any
of these population comparisons are simply probabilities (or
reciprocals of probabilities) that a member of the population in
question would have the stated genotype. There is nothing about
cross-bred individuals, mostly because such individuals do not
generally exist in enough numbers to constitute a recognized
population of their own. The important point here is that "high"
numbers for two separate populations cannot be interpreted as an
indication of admixture of those populations in the test subject, and
there are no numbers for comparison for the hypothetical mixed
populations, and so you can't say *anything* about someone of mixed
ancestry with this method (unless he or she happens to belong to a
mixed population that *is* included in the database).

Note: hypothetical mixed populations can easily be simulated, but
the correction factor for population size is all the more important
for such "fakes" and is uniformly "roughly zero" for all such.

John Chandler


John,

My example maybe the more "exotic," but looking through OmniPop's list of
about 200 populations and based on my own experience, I can conclude that
the tool is interesting for forensics, but has significant limits.

As you allude, it is best suited to relating an individual of ONE particular
population to a list of known populations with corresponding probabilities.
I tested this with a CODIS marker values I know belongs to a contact from an
ISOLATED population. OmniPop worked well in this situation.

However, looking back at my own family history, and probably similar to many
American families; we can trace our lines to many different populations from
around the world, be it in different ratios... e.g. English, Irish, Scot,
French, German, Swedish, Italian, Russian, Jewish, many different
populations in Africa, many different populations in Asia, etc...

What if an individual has grandparents that are: Irish, Italian, German, and
French? For the forensic anthropologist, how will they know if a particular
DNA sample is from a person of ONE particular population, or a person of
MIXED heritage prior to running the tool?

OmniPop has the clever solution of having both "super-populations" as well
as "sub-populations." A "super-population" maybe like "FBI Caucasian" or
"FBI African-American" and "sub-populations" being "Virginia
African-American" or "New York African-American."

Still, the testee needs to belong to ONE population to work well. If not,
it may result in a stastical rank-list that has no relevance for the
genealogist/researcher that is trying to find more clues...


Jim


This thread: