October 11, 2010

Running EURO-DNA-CALC on GenomesUnzipped

(Last Update Oct 23)

genomesunzipped is a new initiative to put data of personal genomics customers online. It's a great idea, and the data will be quite useful to many people.

I downloaded the available data and ran EURO-DNA-CALC on them. Of course it is meant to be used for European or West Eurasian people, which all of them seem to be.

Here are the results for the 12 people whose data were online as of this writing. In bold are components whose confidence intervals do not intersect 0.


Most of them seem to be of NW European descent as their names suggest, and a couple seem to be partly or significantly of Jewish descent.

If you are one of these people, feel free to write to me or leave a comment to tell me if I'm right or dead wrong!

UPDATE (Oct 14): A more detailed analysis of DBV001 and VXP001 in this post.
UPDATE (Oct 23): A much more detailed analysis of all Genomes Unzipped individuals in the context of western Eurasia.
UPDATE (Nov 1): Joe Pickrell discovers Jewish great-grandparent

31 comments:

Vincent Plagnol said...

On my end, I am not aware of any Jewish Ashkenazi background. I'd be curious to figure out where this signal is coming from and whether it is spurious or not.

Vincent Plagnol

Daniel said...

Nice work!

The results fit with expectations overall (Dan is indeed Ashkenazi Jewish, and Luke has some Eastern European ancestry). It will be interesting to see what Vincent and Don make of their results - and we'll also be able to compare them to our PCA clustering shortly (still finalising our reference data-set).

Dan said...

100% correct in my case. To the best of my knowledge both my maternal and paternal family lines are of Ashkenazi Jewish descent.

- Dan

Dienekes said...

On my end, I am not aware of any Jewish Ashkenazi background. I'd be curious to figure out where this signal is coming from and whether it is spurious or not.

I've found out from people that have communicated to me their results that Jews invariably score high on the AJ component of this test (100% scores like Dan's are not uncommon), but low AJ scores on non-Jews also occur.

It could be anything from other Jewish-like ancestry not represented in the test (e.g., some Arabs get high AJ scores), or cryptic ancestry not known to the person, or the limitations of the test itself (as there are 192 SNPs in common between the 23andMe results and the Price et al. panel it is using).

Marnie said...

FYI: Pickrell is a Welsh/Scottish name. Barrett is common in Scotland. McArthur originates in either Scotland or Ireland.

Wright, Morley, Conrad, Anderson and Fisher are not Scottish names.

p-ter said...
This comment has been removed by the author.
Comment said...

There's another test:
For those who wish to double-check, they can participate in Polako's BGA:
http://bga101.blogspot.com/

There already is one Greek participant Dienekes...

Incidentally, a control test for Euro-DNA-Calc is the Ancestry Finder tool in 23andme; if one has up to a 45% Euro DNA Calc score, but 0-1% Ashkenazi Jewish on Ancestry Finder then Euro DNA Calc indicates shared Mediterranean ancestry, not recent Jewish ancestry.

Spy said...

Blogger just munched what I had written (Grrr).

Might EDC have a hard time distinguishing between minority components, especially adjacent ones?

For example, Luke's 61% has high confidence, but his 39% is only good as a combined AJ-SE figure.

Contrariwise, a Jew with 25%NW ancestry may well be 16%SE and 9%NW.

Dienekes said...

I'll post tomorrow or the day after an analysis of Near Eastern populations with respect to Europeans and Africans which identifies a specific cluster centered on Arabs.

As recent studies on Jews have shown shared autosomal ancestry between Jews and Arabs, Jews ought to register -in principle- a relatively high value in that "Arab cluster", while the Tuscans register only 0.3% of it.

It will be interesting to see what DBV001 and VXP001 score in that component, and I'm sure I'll get around to checking that.

Ponto said...

How accurate is EURO-DNA-CALC? It did pick out the sole Ashkenazim member of that genomesunzipped group.

I used Dienekes program some while ago and it came out 87% SE European with 8% NW European and 5% Ashkenazim. The latter figures did intersect zero. I should be mostly SE European being Maltese going back many hundreds of years with no other ancestry. However, other tests, McDonald's for instance, show a significant Middle Eastern score similar to that found in South Italians. I am just wondering why this Middle Eastern score was not picked up by Dienekes program as Ashkenazi.

Marnie said...

Dienekes:

Your August comment on the Myres paper:

"It is interesting to see where R-M269(xL23) is concentrated. In Europe I see cases in Germany, Switzerland, Slovenia, Poland, Hungary, Russia, the Ukraine."

My comment [blog comments, Myres paper, August]:

"Yes, and Western Europe, especially Scotland and Ireland."

Is it possible that the "Ashkenazi" that Euro-DNA-Calc is picking up is instead more broadly the autosomal signature that accompanies populations with an elevated level of yDNA R-M269(xL23) or R-M269(xM412)?

You've mentioned before that there is a strong correlation between J2 and R1b. Maybe that's not any R1b, but a specific type of R1b, such as R-M269(xL23) or R-M269(xM412).

There are three other things that lead me in this direction:

1. Gioiello, who says he is not Jewish, and doesn't know of any Jewish ancestors, is M269(xL23), (R1b1b2a),

2. According to the Myres paper, there is 27% frequency of L23(xM412) in the Upper Rhone Valley.

3. There is the possibility that the name Plagnol or Planiol originates in the Auvergne, which, if I remember correctly, has an elevated level of R-M269(xM412). Vincent, perhaps you could help us here?

Thanks for your consideration.

Dienekes said...

You've mentioned before that there is a strong correlation between J2 and R1b. Maybe that's not any R1b, but a specific type of R1b, such as R-M269(xL23) or R-M269(xM412).

A local correlation in the Balkans, not a global correlation.

Fanty said...

Me in 3 different tests for comparation:

In EURO CALC:
Northwest European likelyhood: 28%
Southeast European likelyhood: 72%

In FTDNA Population Finder BETA:
Western European (reference Populations: Orcadians, French, French Basque, Spanish): 85%
European (unable to pick a group): 15%

In Polakos Project with latest set up:
(these are components as seen in admixture various calculations)

East-Northeast Component: 43.4%
West-Northwest Component: 31.6%
Mediteranean Component: 20.3%
Caucasus/Anatolian Component: 4%
Middle Eastern Component: 0.7%
North-West-African Component: 0%

All 3 tests used the same Raw Data File.

Gioiello said...

Marnie, not only I have a paper trail that links me to an ancestor of mine of the beginning of the 14th century (Signorino del Badia) who lived in Castelfiorentino near Florence, but I am also K1a1b1, the ancestor of the most part of Ashkenazi Jews who are K1a1b1a, but my mutation 9932A is probably typically Italian from very ancient times. Also a relative of mine, my cousin’s husband Fabrizio Federighi, tested by me at SMGF, is R1b1b2* (L23-), very close to many Near Eastern haplotypes, and maternally R0a, but with mutations we find only in Tuscany. I am not able to explain my 23% and 19% of Ashkenazi ancestry by 23andMe and deCODEme. In fact there isn’t a recent relatedness with Jews: on Relative Finder at 23andME I have a few persons linked to me, and all of European (mostly from the British Isles) descent, then the link with Jews isn’t recent.

I permit not agreeing with Dienekes about the Balkan origin of R-L23 and R-M412/L51: I think having demonstrated ad abundantiam that their origin is in Italy.

Dienekes said...

Me in 3 different tests for comparation:

In EURO CALC:


FF and EURO-DNA-CALC have only 63 SNPs in common (vs. 192 for 23andMe) so I am highly skeptical of it. I only added it because people were asking me to.

Marnie said...

"A local correlation in the Balkans, not a global correlation."

Yes, of course. But there is likely a correlation between populations with R-M269(xL23) or R-M269(xM412) and Balkan R1b.

Perhaps the strongest correlations are to be found in the Haute Rhone and the Auvergne, as the Myres paper would indicate.

Thanks, Gioiello, for further information.

And thanks, Dienekes, for these Euro DNA Calc runs and to the genomesupzipped team.

pconroy said...

Marnie,

The name Barrett is Irish, see more here:
http://www.ancestry.com/facts/Barrett-places-origin.ashx

http://www.irishtimes.com/ancestor/surname/index.cfm?fuseaction=Go.&Surname=Barrett

pconroy said...

Marnie,

The name McArthur is Scottish:
http://www.ancestry.com/facts/McArthur-places-origin.ashx

http://www.irishtimes.com/ancestor/surname/index.cfm?fuseaction=Go.&Surname=McArthur

Though of Gaelic descent, from the Celtic Scotti.

Marnie said...

Thanks, pconroy, for checking on "Barrett". So more likely Irish, then English, then Scottish. Thanks for checking.

A variant of the name, Baret, appears on the 1296 Ragman Roll, an early definitive document of Scottish landowners. The name is recorded in county Pebbles (now county Tweedale.)

It's common to see multiple spelling variants for old Scottish names.

Whatever its ancient distribution, it's clear that today the name is most common in Ireland.

Again, thank you for looking this up.

mcdonald said...

I ran these people through my "locate on the map" program, and I get the folloeing results:

All are British except:

DBV: Jewish (or Greek, but probably Jewish).

JKP: SwissBavarian/Austrian

JXA: Low Countries or western German.

VXP: French.

However, the bottom three could also be some mix which averages out at that spot.

Doug McDonald

Jan Aerts said...

@Doug: Could you elaborate on your "locate on the map" program? I (JXA) am Belgian...

Marnie said...

Gioiello:

Do you know how many SNPs 23andMe and deCODEme used to get your results?

mcdonald said...

My "spot on the map" does a PCA
on a person (using 70 comparison
populations) and then calculates the average value of the first 9
principal components of each of the 70 populations. It then least squares fits the test person to those 70 averages. It usually finds
many acceptable fit combos. It then calculates the latitude and longitude of the weighted averages of the comparison population that make up the best
few (2 to 20) fits to the test person.

That's the short, elegant, description. The long description is like the inside of a sausage factory.

Doug

Marnie said...

Doug,

Conrad is originally a German or perhaps Czech name.

I wasn't originally sure about Aerts name, but when you look at his picture, you can be pretty sure that he is German or from the Low Countries.

Joe Pickrell seems to be the stand out. His name (Welsh or Scottish) doesn't match your analysis. It would be interesting to know why.

I have to say that until these tests can be definitive, I'm not compelled to entirely believe them.

If I were to take such a test, I'd probably be told something that is misleading, or something that I already know. The margin of error isn't narrow enough to tell me something new, such as when my ancestors reached the British Isles or clearly indicate how much Norse, Irish, English and Scottish I have in me.

bear of little brain said...

ilana fisher sounds like a very jewish name to me - but scored zero in the ashkenazi jewish column.

Fanty said...

"Conrad is originally a German or perhaps Czech name"

Conrad is Germanic.At least it originates in the Germanic words
Kuoni = keen
rat = advisor

Distribution of the name in Germany:
http://verwandt.de.karten.s3.amazonaws.com/conrad_relative.png

Distribution of the name in France:
http://familleunie.fr.cartes.s3.amazonaws.com/conrad_relative.png

Gioiello said...

Marnie writes:"Gioiello:Do you know how many SNPs 23andMe and deCODEme used to get your results?"

I remember that Warwick ran my raw data on the Dienekes' Calculator (we were at the end of 2008), then I think Dienekes can replay to you better than me.

Marnie said...

Gioiello,

I just checked the 23andme website. They say they return 550K SNPs.

Dienekes used 275K SNPs to resolve Dan Vorhaus' Ashkenazi background and Vincent Plagnol's Western European background.

I suspect that 23andme wasn't returning 550K SNP's in 2008.

Have you contacted 23andme? Why don't you ask them to redo your test with the most updated number of SNPs? And of course they should do it for free.

And after running with the latest number of SNPs, if there is still an error, 23andme should be made aware of it and should be telling people with this particular genetic background that there may be an error in evaluating the degree of their Ashkenazi ancestry.

Beyond that, it may take quite specific SNPs to resolve your particular Italian ancestry from someone with an Ashkenazi background. It may not even be possible. I have friend who is of an ancient Jewish Slovenian family. He has red hair and freckles and looks very Western European.

Again, thanks Gioiello, for telling us your story.

Gioiello said...

Marnie, certainly my 23andMe has 550k SNPs. I had also V3 for my Y (see "Adriano Squecco's spreadsheet"). Unfortunately I lost my Office on my PC by a virus and a can't now load my raw data, but if Dienekes wants I can give him my username and password both for 23andME and deCODEme and he could run again my results.

Anyway my thought has always been that are Jews (above all Ashkenazim and Sephardim)to be Italian (or European) and not the contrary.

Marnie said...

Gioiello,

You can probably take your PC to an anti-virus and spy-ware shop and get them to clean up your PC. That costs about $100 here in the Bay area. I am not sure how much it costs in Italy.

I don't want to get in the way of your idea about an Italian refugium for R-M269(xL23). Groups like Myres et. al. are looking at this, I am sure.

To look at that clearly, teams will have to look across multiple geographically located samples.

In the meantime, it is up to you if you want to re-examine your data. It would be helpful for everyone to know if your genetic ancestry can be resolved into an identifiable Northern Italian signature.

Have a nice weekend.

Gioiello said...

Marnie, yesterday I sent my access keys to 23andMe and deCODEme to Dienekes. I don't know if he wants to run my data again. I am open to the truth and every result.