Archiver > GENEALOGY-DNA > 2002-08 > 1030558442

From: lmpajg <>
Subject: [DNA] How Specific is 12 marker test
Date: Wed, 28 Aug 2002 20:14:45 +0200

This is a multi-part message in MIME format.
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

Content-Type: text/plain;
name="Draft Results .txt"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline;
filename="Draft Results .txt"

Over the last few weeks, there has been a lot of discussion on this list =
about the value of running a Y-chromosome surname study with a smaller nu=
mber of markers before deciding on an upgrade to a larger set of markers =
if testees match. FamilyTree DNA offers a less expensive test at 12 loci=
, followed by an optional upgrade to 25 loci at an additional price. Upg=
rading to a larger set of markers provides a shorter estimated time to mo=
st recent common ancestor (MRCA) for subjects who match closely, but it a=
lso carries a higher probability of not matching as well on all additiona=
l markers. Some members of this list believe that a smaller set of 12 ma=
rkers are "next to worthless" in sorting out family relationships, while =
others have stated that 12 markers worked fine for their studies. =20

I decided to look at this question numerically, using the evidence from t=
he surname studies to date that have 25-marker results: what became of t=
he 12-marker matches when expanded to 25 markers?

Data here are from the 16 surname studies on Chris Pomeroy's and Kevin Du=
erinck's lists of Y-chromosome surname studies which had posted on the we=
b detailed results of 25-marker tests for at least two individuals as of =
August 15, 2002. Excluded were a number of studies with only one testee =
reporting 25-marker results so far, and many others with studies in progr=
ess but no results posted yet. The 16 studies covered here include 154 i=
ndividuals tested on 25 loci, in addition to 186 subjects with only 12-lo=
ci results posted. =20

Since many of these testees made a conscious upgrade from 12 to 25 marker=
s, there's a built-in bias to high initial matches in this set - more 12/=
12 and 11/12 matches than you would probably find on average in a surname=
study. However, not all of these 154 individuals had gone through a two=
-step process of testing 12 markers first, then upgrading to 25 if their =
matches were close - some surname studies had some or all of their partic=
ipants tested initially at all 25 loci. For these, I tallied results bas=
ed on the subset of 12 markers as well as on the whole 25. All compariso=
ns were made strictly between individuals within a surname study - no com=
parisons of individuals across studies. And to limit the number of pairw=
ise comparisons, I selected the best match for each individual based on h=
is 25-loci results, then tabulated where his 12-loci results would have p=
laced the match. The work was greatly simplified by the fact that most s=
urname study coordinators have presented their results grouped by strengt=
h of the match.

Results are as follows: =20

Overall most of the matches found to be "close relatives" at 12 markers t=
urned out to be still close at 25 markers - defining "close relatives" as=
matching at least 11 out of 12 markers or on at least 23 out of 25 marke=
rs. But weaker matches, especially at 9 or fewer out of 12 markers, were=
even weaker at 25 markers:=20

Matches----------------Matches out of 25--------------------Row
out of 12____25__24__23__22__21__20 or fewer___Total

9 or fewer___na__na__na___0___2____31_________33

Col .Total___60__30___9___4___6____45________154

* Ex: '60 of the 90 matches at 12 out of 12 markers also matched at 25 o=
ut of 25'

Rounding the results off to simple numbers, and on the conservative side,=

* For matches at 12/12 markers, in more than 4 out of every 5 cases (78/9=
0), the relationship still remains close, at least a 23/25 match, if 25 m=
arkers are tested
* For matches at 11/12 markers, in more than 2 out of every 3 cases (20/2=
7), the relationship holds to at least a 23/25 match.
* For matches at 10/12 markers, only 1 out of 4 reached a 23/25 match; mo=
st were at 21/25 or fewer.
* For matches at only 9 or fewer markers, the best accomplished was a 21/=
25 match; most were at 20/25 or fewer.

These results show that 12 markers can take you quite a ways in sorting o=
ut relationships - definitely 12 are enough to separate out cases who are=
not closely related (the cases of 9/12 or fewer matches), or to suggest =
that a relationship may be borderline (the 10/12 cases, if not otherwise =
supported). As others on the list have noted, 12 markers are definitely=
enough to sort out people who *don't* connect. =20

On the other end of the spectrum, if a match is 11/12 or 12/12, these dat=
a suggest that the majority of these individuals will remain closely rela=
ted even at the shorter time to MRCA that 25 markers provide. In fact, w=
ell over half of those who had zero or only one mutations between them on=
12 markers remained at the same zero or one mutation on all 25 markers (=
cases on the diagonal in the matrix above). Finding such a relationship =
at the lower price for a 12-marker test sounds very attractive. But the =
exceptions are sobering - about 1 out of every 5 perfect 12/12 matches an=
d about 1 out of every 3 of the 11/12 matches would not turn out to be cl=
ose relations at all after testing at 25 loci. To clarify these matches=
, a larger set of loci was essential. =20

I would be interested in hearing what other list members think about thes=
e results. I personally was surprised that the additional marker upgrade=
didn't lead to even finer discrimination among cases - more people staye=
d "closely related" at the 25-marker step than I might have expected befo=
re looking at the data. (Perhaps it was just good old-fashioned paper ge=
nealogy work that helped these testees make good decisions on who/when to=
upgrade?) =20

Thanks to all those who have made their numerical results available on th=
e web for exercises such as this.



This thread: