April 13, 2008

Ashkenazi genes

On GNXP.com, Greg Cochran points to an interesting graph of the genetic distribution of individuals of European descent along two "principle component" axes:
"This SNP study (and others) also shows that Ashkenazim are genetically distinct from other Europeans, which allows fairly accurate identification of group membership. Almost perfectly distinct, if you look at Ashkenazim whose grandparents are all Ashkenazi (the violet dots)."

The violet dots that cluster in the upper right amidst purple dots are Ashkenazi [Northern European] Jews with four Ashkenazi grandparents. The more scattered purple dots are self-identified Ashkenazi Jews with more mixed pedigrees. The green dots in the upper left corner are Irish (this may be the first racial graph I've seen in years where the color coding wasn't intentionally randomized) and the light blue dots at the bottom are Greeks.

I don't know what the two axes are -- factor analysis is a creative tool, in which you've got to come up with the insights. The statistical mechanics won't do it for you.

My published articles are archived at iSteve.com -- Steve Sailer

20 comments:

Anonymous said...

What, the x-axis is guile and the y-axis is intelligence?

Anonymous said...

to a first approximation looks like Y = N-S; X = E-W

Anonymous said...

Yeah, I mean, it's gonna be historical. Relating it to personality traits might work out but what you've probably got is a cladogram where groups that diverged earlier will be further away.

It's also worth mentioning that just because two groups have similar origins doesn't mean they're similar now; leaving aside the Ashkenazi (who have been displaced far and subjected to some unusual selection pressures), look at the difference between next-door groups such as the British and the Irish, or the French and the Germans.

Anonymous said...

The original GNXP piece also dealt with some esoteric but important details of Ashkenazi Jewish genetic history. In particular a relatively recent, massive data set showed that Ashkenazi Jews have existed as a definable genetic group for an extended time, and that as a group they have not been subject to either severe 'bottlenecks' or recent 'founder effects.'

This new work also identified regions of the Ashkenazi Jewish genome that have been under recent positive selection. As Cochran notes, with a touch of 'I told you so,' among these regions are segments containing genes associated with genetic diseases such as Tay-Sachs and other sphyngomyelin metabolism defects which are much more common among Ashkenazi Jews.

Cochran and two others have previously proposed, in a controversial paper, a possible reason why these genes have been strongly selected for in the face of the highly deleterious consequences for people with two copies of the mutant gene.

Anonymous said...

very clever, let's!

Anonymous said...

The axes look like they come from principal components analysis, which would mean that the axes mean.... NOTHING.

That is, you can't interpret it as east/west; north/south; tall/short. The point of PCA is to make clusters, to show distinctions between groups. The PCA was being applied to single nucleotide stats, and it just shows these groups are genetically well-defined (that certain groups share certain patterns of SNPs). That's all.

Anonymous said...

I wonder what Alicia "A.K." Keys would make of all this? The Ashkenazi jews who own the record companies and have aggressively marketed gangsta rap are "white" when theyre being slammed in the press--imagine A.K. talkin' bout them jews whats pushin' all this gangsta rap,y'all,---but "change back" when the doors are closed!

Anonymous said...

This chart in isolation implies that Jews are very different from other Europeans, but that is false, it's only saying that Jews have maintained a separate breeding pool for about 1500 years or so.

If you were to place any other non-European group into the chart, such as actual people from the Middle East, they'd be very disimilar to Ashkenazim.

The vast majority of Ashkenazim genes trace back to Caucasian Rome.

Anonymous said...

Actually, I'm pretty sure that Half-Sigma is mistaken. The figures I've seen are that somewhat over half of Ashk Jewish ancestry is Middle-Eastern and somewhat under half is European. This is pretty much what you'd anyway expect based on extremely crude estimates using "marker genes" such as those for blue eyes.

As I've said elsewhere, it's too bad that the chart didn't include a sample of Middle-Easterners/Levantines. It would be interesting to see whether the Ashk Jews did fall roughly half-way between that group and the Europeans.

Anonymous said...

HS is wrong.

An earlier study finds that Armenians cluster with AJs.

HS is engaging in wishful thinking at best, willful deception at worst.

Nor do genetic studies support the claim that the "vast majority of Ashkenazim genes trace back to Caucasian Rome", unless you construe "Caucasian Rome" to include Middle Eastern immigrants.

Unknown said...

I'd like to see Chinese, Thai, Serbs, Africans, Egyptians on the same plot to get some additional idea as to whether Jews are "close" to Europeans or not. I would certainly expect them to be, but would like to see it.

Anonymous said...

Unless Hungary, Ukraine, Austria, Romania, etc have suddenly moved without telling anyone I doubt whether 'Northern European Jews' is a good way to describe Ashkenazi Jews.

nomen said...

There is a paper by Bauchet et al. from 2007 that includes autosomal SNP data from small numbers of individuals across a wide range of populations. This is the paper that shows Jews clustering with Armenians, Greeks, and some Italians. There are also a very small number of Middle Eastern and North African individuals included in the dataset who are outliers from that cluster. It would be nice to have more individuals and more SNP's than were tested in this dataset, but the paper still gives a broad outline.

The paper* can be accessed for free at the following link: http://www.pubmedcentral.nih.gov/articlerender.fcgi
?tool=pubmed&pubmedid=17436249

*Am J Hum Genet. 2007 May;80(5):948-56. Epub 2007 Mar 22.

Mike said...

Tough question, obviously. I think the obvious first step would be to see how each axis (+under various rotations) correlates with historical ethnic latitude, and time since adopting an agricultural lifestyle.

Anonymous said...

Steve,

Does that mean you now accept the advantages of cluster analysis (PCA) over "partly inbred extended family" as a way of understanding "race"?

AJs and Europeans are certainly all part of the same extended family, but precisely how different, or similar, is revealed by the clustering.

Re: AJs clustering with Armenians or middle easterners, I believe that is only the case when a small number of SNPs or loci are used. As the number of loci gets larger the resolution gets much better and AJs will almost certainly resolve from those other groups.

Anonymous said...

I'd like to see Chinese, Thai, Serbs, Africans, Egyptians on the same plot to get some additional idea as to whether Jews are "close" to Europeans or not. I would certainly expect them to be, but would like to see it.

Well, we get a non-trivial amount of info on raw distance just by eyeballing the graph; AJs are as far from Italians and Irish as Italians and Irish are from one another, about ten times as far from Germans and Irish than either is from one another, etc.

But I'm over my head with this stuff, so I have no way of knowing what these differences really amount to. And yeah, I'd like to see a zoom-out too.

Anonymous said...

Jews look like middle easterners, not Europeans. A lot of people don't realize the northern "Arabs" dressed in Western clothing without beards look more european than they think. Jews look like a cross between northern Arabs and people from the Cuacuses. They don't look at all like people from northern or western Europe.

Anonymous said...

HS is engaging in wishful thinking at best, willful deception at worst.

Looks like the new chart that Steve posted proved that I was correct, and you are wrong.

Anonymous said...

An unpublished analysis done using a tiny handful of likely selection-prone SNPs does not trump analyses using 10,000 to hundreds of thousands of SNPs. The fact that you so readily accept the conclusion of an analysis whose methods you know nothing about (when you like the results) suggests you are trying to push a predetermined line and have little regard for the truth.

Anonymous said...

"The fact that you so readily accept the conclusion of an analysis whose methods you know nothing about (when you like the results) suggests you are trying to push a predetermined line and have little regard for the truth."

Could someone kind of spell this out for the neophyte? What predetermined line prefers the idea that AJs are like Europeans to the idea that AJ are like Arabs? This is all a bunch heavily-crunched numbers on gene frequencies, right? Why does it matter politically? My guess is that the arguments on this thread are between people of different ideologies but I could only guess (coin-toss style) which ideologies liked which ethnological assertion.