GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2009-10 > 1256138372


From: David Faux <>
Subject: Re: [DNA] Inconsistencies in 23andMe Ancestry Painting
Date: Wed, 21 Oct 2009 08:19:32 -0700
References: <ea3bd9560910201821h1fd9b428m83f69c2a20e07d50@mail.gmail.com>
In-Reply-To: <ea3bd9560910201821h1fd9b428m83f69c2a20e07d50@mail.gmail.com>


By the way I want to make it perfectly clear that I am not leveling any
criticism at 23andMe, only offering some observations that they might want
to use to "buff up" their present offering in terms of the algorithm. I am
well aware that the determing factor is "resource allocation", but I will
continue to advocate for those features that I believe many of us would
appreciate.

The Ancestry Painting, argueably the most important ancestry feature of the
entire test, but it has not been revised since April of 2008. A great deal
has happened since then in just the published literature. I have been in
contact with three of the scientists at 23andMe and they fully sympathize
with the concerns I and others have been proposing, but unless people
agitate to have this amended or updated it may sit indefinitely as they move
on to other topics which are perceived as of greater import.

As I await the arrival of my uncle's results, I am becoming increasingly
"concerned" that the X chromosome has, after all this time still not been
painted. I keep repeating these concerns since that is simply the way the
world works - the team at 23andMe needs to know what the customers wish to
see. Many of us are scientists and know what is feasible and are not asking
for a trip to the moon. FTDNA is fantastic in listening to customer wish
lists. 23andMe is rather new on the scene and is getting a feel for who is
out there in customerland. I think it is safe to say that a large number
are unwilling to accept the use of groups such as the Han as proxies to
Native American - for example. Apples to apples are the only valid
comparisons.

The good news is that the new Relative Finder will include comparisons of
the X chromosome. This is a wonderful feature that will a little fine
tuning will provide a unique opportunity to explore relatedness from an
angle never before viewed by the public. Kudos to 23andMe. I am very
anxious to see the data from my uncle and various levels of known cousins
using this finding tool. I anticipate being able to identify blocks from
Canadian Loyalist Palatine German ancestors due to multiple cousin marriages
that occurred up until quite recently, and a lot of cousins who have tested.

David K. Faux.

On Tue, Oct 20, 2009 at 6:21 PM, David Faux <> wrote:

> List,
>
> For some time I have been examining the Ancestry Paintings of the people
> who I share findings with - over 100. Most are 100% European, but perhaps a
> third have some degree of "Asian" and / or African blocks showing in their
> "painting".
>
> Something that was said in another forum prompted me to go to the painting
> of Dr. Roy King (a prominent American of African - American heritage) and
> examine each chromosome closely. The painting and interpretation is given
> here:
>
> http://spittoon.23andme.com/2008/06/13/a-beautiful-ancestry-painting/
>
> What is odd (and Dr. Ann Turner picked up on this before I noted it) is
> that many if not most of the large number of Asian blocks extend the full
> width of the chromosome. This makes absolutely no sense from any
> genetically oriented explanation I can come up with. Even fully siblings
> would not show this degree of correspondence (or exact placement of small
> haploblocks). Some of these blocks appeear to be over 10 Mb and others
> cannot be over 1 Mb - but how can they extend beyond the half way point on
> any of the chromosomes. Something is clearly amiss.
>
> The second factor is that on any person with majority European ancestry
> that I have seen, there will be some (e.g., two) blocks of say 5 Mb (it is
> impossible to tell the exact size since the dimensions are not given nor are
> the start - stop positions). None go over the half way point to extend
> across the width of the chromosome. Why are there very narrow blocks in Dr.
> King's painting, but never in the large number of paintings with majority
> European that I have seen, is any block that small, even remotely so, shown
> - never.
>
> I have said all along that there is a bias inherent in the algorith. On
> the first pass it determines the major parameters of the genomic ancestry
> structure. Then comes the if - then components. Clearly any Asian like
> blocks in someone with African and Asian are allowed through the filter and
> "painted". However if the algorithm has determined "European" a different
> set of criteria appear to be applied - my guess is if less than 10 Mb
> (perhaps 5 Mb - translated into cM equivalent) assume false positive and
> attribute to "European". I am not sure all of the orange is really Asian or
> Native American - but more likely misclassified African - hence the block
> extending across the chromosome. The 23andMe company representative reports
> that these findings are characteristic of the San - the resemblance to East
> Asians (not seen in the Nigerians used as reference samples in 23andMe).
>
> The bottom line is that due to fear of false positive in an otherwise blue
> (European) background it is necessary to have a much larger block than if it
> is found that your ancestry is a mosaic of all three groups. Or so it would
> seem.
>
> I am only giving an opinion here, but there is not a shadow of a doubt in
> my mind that the algorithm 23andMe employs classifies people on the basis
> of criteria that depend on the Gestalt of the genomic structure, and us
> primarily "blue folk" are never going to see the micro blocks that are
> portrayed in Dr. King's painting. However the timely presentation of the
> Browning and Hon team noted in an earlier posting today speak elegantly to
> this issue. Identical by descent can be teased out even at the 1 Mb level -
> but it will be necessary to use a program such as BEAGLE and to broaden the
> reference samples (the Han simply will not cut it as a reference group for
> those of us who wish to find evidence of Native North American ancestry -
> ok, I am a broken record on this point).
>
> David K. Faux.
>

This thread: