View Full Version : Threshold for regional admixture "noise" ?

05-18-2016, 09:44 PM
Is there a standard percentage below which a regional ancestral admixture signal is considered statistical noise?

I'm (at least very nearly) entirely European, but Geno 2.0 Next Gen includes a 2% "Arabian" admixture result for me. FTDNA's MyOrigins apparently puts that same admixture I have as a 4% "Middle Eastern" percentage.

No doubt that even if these are statistically significant, they're just leftover signals from Neolithic agriculturalist expansion into Europe. But I'm wondering if it's even worth paying that much attention to. Are they even meaningful at all?? Or just "placeholder" designations for otherwise unspecified admixtures?

05-19-2016, 10:26 PM
No doubt that even if these are statistically significant, they're just leftover signals from Neolithic agriculturalist expansion into Europe.

Four per cent might be more than a Neolithic remnant. Please post a couple of your Gedmatch oracle results e.g. PuntDNAL K12 Modern, Eurogenes K13, or Dodecad K12b, with four-way mixtures.

05-21-2016, 02:57 AM
I'm not sure you can really find a "standard" percentage. I would think it would vary by company and by calculator, as well as by exactly what regional admixture is being considered. Some components are easier to identify than others.

For example, I have a Native American component that shows up at 2.0% at 23andMe; at 1% at Ancestry; and at 1.9% at DNA.land. It doesn't show up at FTDNA, but a 1% Northeast Asia shows up instead. (And 1% Native America shows up for my brother.)

Obviously this is well under 4%, but it's fairly persistent across a variety of calculators. In addition, I do have a documented Native American ancestor, and numerous fellow descendants of this ancestor appear as cousins on Ancestry. Most of them, like me, show at least "trace" Native American. (And some of them show more, especially those with several lines of descent from the ancestor in question.)

This is against a predominantly European background, and it's a very small percentage -- but I don't believe it's just "noise". (I also show a small -- 2% -- Eastern Middle East percentage at FTDNA, but pretty much nowhere else, so I'm less certain about this one.)

05-21-2016, 09:00 AM
Here's the Eurogenes K13 admixture breakdown (https://www.dropbox.com/s/h69s9peeyy3i1zx/Eurogenes%20K13%20results.jpg?dl=0), as well as the Eurogenes V2 K15 breakdown (https://www.dropbox.com/s/etzfoaynlrt4rou/Eurogenes%20K15%20results.jpg?dl=0).

I don't know if it makes a difference or not, but the raw data came from an older FTDNA test, so this doesn't reflect the new FTDNA MyOrigins results (that use my imported Geno 2.0 Next Generation raw data) to derive a 4% Eastern Middle East admixture.

Help? Thanks!

05-21-2016, 03:23 PM
Thanks. If the results are based on the Affymetrix chip then they are suspect, the error rate was high.

Your West Med is on the high end for English but on the low end for French or German. Essentially, like many Caucasian Americans, you look like you are swimming around in the English Channel.

The more interesting thing in your results is that your Siberian is quite high. It's the sort of level you'd expect to see for a full-on Eastern European or someone from the Caucasus. The West Med could be accounted for by French or German genes but your Siberian is significantly higher than you'd see in Germany unless you had something like a Tatar great-grandparent. That might indicate some ancestry from farther east, but again, the Affymetyrix chip was not entirely reliable.

05-21-2016, 05:53 PM
Thanks so much, I appreciate it. My maternal grandfather was from a Swiss/French line, so that clearly explains the West Med. My Geno 2.0 NG results and revised FTDNA MyOrigins results have me in a similar West Med / Continental context.

I'm at an utter loss to explain the Siberian thing. It's possible that it's real; I have two 5th generation great-grandmothers that I know nothing about, and a good number of 6th generation ancestors who are completely unknown to me (and that ~1+% is just about the right % admixture level for a 5th or 6th generation ancestor). But then, it could just be noise, like the Oceanian almost certainly is.

Any thoughts on the Middle Eastern thing?

Again, thanks so much.

05-22-2016, 02:52 AM
You're most welcome. The Siberian might be noise but OTOH if it seems to show up in multiple assessments and is at significantly above-average levels it might be something worth looking into.

The Middle Eastern looks at first glance like something not real but rather a "redirection" artifact to place you a little southeast of wherever you would otherwise appear. You might just want to run your Gedmatch by-chromosome admixture results, take a note of the locations any Middle Eastern or West Asian painting occurs, and note who matches you there. If you don't have any matches with people with Middle Eastern ancestry, it's most likely a red herring.

You could do the same with Siberian and see if there are any patterns.