PDA

View Full Version : Commercial DNA tests have difficulty determining ancestry



Angela
28-08-16, 17:59
This is a note written by Joe Pickrell of DNALand, who is also a respected geneticist.

What is ancestry?

https://medium.com/@dl1dl1/what-is-ancestry-842109cb8ebd#.th2hby2si

He's honest, but I don't think he goes far enough....

"This suggests that people expect a genetic “ancestry test” to predict the geographic and/or ethnic labels of their ancestors. Unfortunately, if you sit down and try to write an algorithm to do this, you will immediately come across two huge and mostly intractable problems.

Problem #1: What time depth are we talking about?

Obviously we all have ancestors that lived at different times. You had maybe 8 ancestors living 100 years ago, but many thousands that lived 500 years ago. So whose geographic and/or ethnic labels should we try to guess — those of your ancestors living 100 years ago, or those living 500 years ago? (Or 1,000 years ago? Or…?)."

"So it’s not totally clear what time depth people generally think of when they think about their ancestry. Indeed, it seems plausible that the “correct” time depth to report in an ancestry test depends on a user’s…ancestry. This should be a hint that this is not something that can be objectively read from DNA."

"Problem #2: Ancestry identifiers are influenced by social and political factors...

Indeed, construction of a shared ancestral identity was (and remains) a method for consolidation of political power in ver6 diverse cultures (see e.g.Franco in Spain (https://en.wikipedia.org/wiki/Francoist_Spain#Nationalism)[COLOR). This is largely invisible to genetics, except after hundreds or thousands of years (if shared identities influence subsequent marriage and/or migration patterns)."

"what you would ideally like to have is a detailed list of your ancestors at different time depths, each labeled with their geographic location and any ethnic self-identifiers. You could then say, for example, that 100 years ago 25% of your ancestors lived in Illinois and identified as Jewish, while 500 years ago 5% of your ancestors lived in present-day Andalucia and identified as Muslim."

Unfortunately genetic tests are about as useful as Ouija boards for obtaining much of this information, so we’re going to have to compromise with some dramatic approximations [3]. Specifically, the approach taken by all of the commercial companies (and that we take as well) is to try to estimate the general geographic regions where your ancestors lived (and in a select small number of cases their ethnic identifiers) some indeterminate time in that past, probably something like a few hundred years ago."

"But the key is this: if we replace an impossible goal of perfectly understanding the geography and ethnicity of your ancestors with the more realistic goal of getting a general understanding about some of them, we can now make some progress. "

There's other stuff in there; definitely worth a read, but the fact is that if people really understood what he says here how many of them would actually do these tests?

The other huge problem is that a lot of people, either because they don't know these things, or because of some agenda, take the results of things like the gedmatch calculators, which are watered down versions of Admixture and Structure, absolutely literally, and even worse, actually use them to attempt to reconstruct population history on national and regional levels. It's breathtaking in its audacity, especially when done by amateurs who don't even understand how Admixture works.. So, a "Balkan" score in modern people equates to "Greek" ancestry from the Classical period. How on earth could you conclude that?

I also quibble with something he says about 23andme vs. AncestryDNA. He says he tends to think that 23andme shows more "recent" ancestry than AncestraDNA, i.e. is limited to 500 years ago. That's ridiculous in my opinion.I don't know of Ancestry DNA can figure out your ancestry at a greater time depth than five hundred years, but I do know that 23andme does as well, although they don't admit it. Northern Italians don't have 11, 20, 27, whatever, % "NWEuropean" ancestry that entered their genome in the last 500 years, and nor do Southern Italians have additional "Caucasus" ancestry since 1400 AD. So, even that isn't accurate. Plus, upthread he says they all are giving you ancestry within the last 500 years.

I take nothing away from these people in their knowledge of genetics. The problem is that they know next to nothing of European history.

Ed. The real take home is that these things cannot give people the information they really want. It really only works for people who have all four grandparents from one area, and even then all it's going to tell you is if you are roughly similar to other people of your time and place.

Yes, it can also tell you if you are really the child of your parents, and perhaps it could reunite some lost child or some "cousin", with his or her lost family, but that isn't what people are signing up for....

Twilight
29-08-16, 01:07
I'm really sorry about this but you're losing me here with the vent. When did Gedmatch Balkan get associated with the Greco-Romans? Isn't it supper common for Europeans to get Middle Eastern Dna?


Anyways, ancient ancestry is never set in stone as well as archeology + linguistics. If ancestry percentage is what really attracts people to Dna testing, then I think that's a good thing (in the long run). The more DNA testers, the more data we can possibly use to make our DNA tests more accurate; perhaps we can also add some ancient Civilization samples; for example comparing actual Greco-Roman or Bronze Age Britain/Irish DNA to DNA companie's samples for fact checking. Its just a thought going foreword. We can also use history to educate people on what the they are possibly seeing and show disclaimers saying that it might not be 100% accurate; same goes to archeology.

Fire Haired14
29-08-16, 04:28
23andMe does a good job determining regional ancestry. You're bringing up good negatives in this thread but exaggerating them. All results I've seen are consistent with genealogy.

Angela
29-08-16, 05:38
Maybe I should go back and highlight or put into italics all Joe Pickrell's comments. That was the majority of the post, so I guess he was venting as well? He apparently knows less about these methods and what they can and cannot do than our members? Good to know.

I can't explain it any better than he does, I'm sure, so I would suggest you go back and read his entire article. These commercial ancestry tests can't give people the information they seek, in most cases.

Neither can the gedmatch calculators. I don't care how many modern samples you collect. By the way, you don't need thousands of samples to figure out modern autosomal structure in a given region. A couple of hundred is usually enough for most areas, as Razib Khan pointed out in a recent blog post, so long as they're the right samples. It's a common mistake among amateurs (including science writers) to think that if you get 10,000 people from a certain relatively stable area the autosomal results will be a lot different than if you have a hundred. It doesn't work that way for autosomal dna, just as it doesn't work that way for most statistical models, i.e. you don't need to test a huge percentage of the target population. It's called statistical sampling, right? The important thing is that the samples have to be properly chosen for the area taking into account what we know about the history of the different migrations (so more samples will have to be taken in more areas in Italy than in Denmark, for example), that the samples have all four grandparents from the same area, and, of course, that they are randomly chosen. It can't be a self-selected sample like those from commercial testing. They have to be properly chosen and verified academic samples by people who have a career to lose if they don't follow proper protocols.

Even then, the calculator may not be able to tell you what % French vs German you are, because there just might not be enough difference between someone from Alsace and someone from Burgundy.

As for using gedmatch calculators or commercial testing analysis to draw conclusions about historical migrations, Pickrell didn't address that, although I did. It's not very reliable even in the hands of experts who understand the limitations of Admixture/Structure, much less in the hands of amateurs. To repeat my point, you can't, for example, use "Balkan" ancestry in modern Italians as a stand in for how much Classical era Greek admixture they have, but people are indeed doing it. I argued with several proponents over on 23andme, and I just recently saw the same point being made on anthrogenica. It's unbelievable.

It may someday be different, if we have enough ancient samples, but you'd have to have quite a few from every historical era group.


"Isn't it supper common for Europeans to get Middle Eastern Dna?

Why would you raise this? I don't think I said anything about it. However, since you raise the subject, "Middle Eastern" Dna is actually hidden in a lot of commercial testing and in gedmtach calculators. There's "Middle Eastern" ancestry, if by that you mean ancestry from people from the Middle East in not only the Mediterranean numbers, or West Asian numbers, but in the Northern European numbers, the Eastern European numbers, the Southwestern European numbers and on and on. If you really reduce things to their essentials, the average amount of such ancestry in Europeans (nations weighted by population), counting both the Anatolian farmers and the "Caucasus like" ancestry, is about 60%. Central and northwest Europe is close to 50%, Southern Europe quite a bit more, Baltic Europe less. So, yes, these calculators, even when they're not wrong, can be very easily misinterpreted.

As to ancient samples, yes, they help, but I'm certainly not convinced that people know how to properly create calculators using them, nor that people know how to interpret them. The DNA Land "analysis" of ancient samples is just the latest example. Faced with an ancient genome, the algorithm does the best it can, and assigns to groups with a lot of that ancestry, but faced with all the formal testing that's been done, is someone really going to claim that modern Slavs are 100% ancient Yamnaya or Corded Ware?

A little knowledge can be a dangerous thing.

@Fire-Haired,

No, it doesn't, not even for me, and I'm the best case scenario because all four of my ancestors are from one place. It tells me I have X percentage of British/French, and Balkan, which according to them came to Italy within the last 500 years. I have almost all my lines traced back to the early 1500s, and no, there was no gene flow into me from those areas since that time. If it was accurate at saying where one's ancestors have been since 500 years ago, it would tell me I'm 100% Italian. As for admixed individuals like Americans, who are the bulk of their customer base, it can't tell them what they want to know, which is what % Irish, and what % English, and what % Scottish, and what % German, or Dutch, or French they are.

Twilight
29-08-16, 06:38
Maybe I should go back and highlight or put into italics all Joe Pickrell's comments. That was the majority of the post, so I guess he was venting as well? He apparently knows less about these methods and what they can and cannot do than our members? Good to know.

I can't explain it any better than he does, I'm sure, so I would suggest you go back and read his entire article. These commercial ancestry tests can't give people the information they seek, in most cases.

Neither can the gedmatch calculators. I don't care how many modern samples you collect. By the way, you don't need thousands of samples to figure out modern autosomal structure in a given region. A couple of hundred is usually enough for most areas, as Razib Khan pointed out in a recent blog post, so long as they're the right samples. It's a common mistake among amateurs (including science writers) to think that if you get 10,000 people from a certain relatively stable area the autosomal results will be a lot different than if you have a hundred. It doesn't work that way for autosomal dna, just as it doesn't work that way for most statistical models, i.e. you don't need to test a huge percentage of the target population. It's called statistical sampling, right? The important thing is that the samples have to be properly chosen for the area taking into account what we know about the history of the different migrations (so more samples will have to be taken in more areas in Italy than in Denmark, for example), that the samples have all four grandparents from the same area, and, of course, that they are randomly chosen. It can't be a self-selected sample like those from commercial testing. They have to be properly chosen and verified academic samples by people who have a career to lose if they don't follow proper protocols.

Even then, the calculator may not be able to tell you what % French vs German you are, because there just might not be enough difference between someone from Alsace and someone from Burgundy.

As for using gedmatch calculators or commercial testing analysis to draw conclusions about historical migrations, Pickrell didn't address that, although I did. It's not very reliable even in the hands of experts who understand the limitations of Admixture/Structure, much less in the hands of amateurs. To repeat my point, you can't, for example, use "Balkan" ancestry in modern Italians as a stand in for how much Classical era Greek admixture they have, but people are indeed doing it. I argued with several proponents over on 23andme, and I just recently saw the same point being made on anthrogenica. It's unbelievable.

It may someday be different, if we have enough ancient samples, but you'd have to have quite a few from every historical era group.



Why would you raise this? I don't think I said anything about it. However, since you raise the subject, "Middle Eastern" Dna is actually hidden in a lot of commercial testing and in gedmtach calculators. There's "Middle Eastern" ancestry, if by that you mean ancestry from people from the Middle East in not only the Mediterranean numbers, or West Asian numbers, but in the Northern European numbers, the Eastern European numbers, the Southwestern European numbers and on and on. If you really reduce things to their essentials, the average amount of such ancestry in Europeans (nations weighted by population), counting both the Anatolian farmers and the "Caucasus like" ancestry, is about 60%. Central and northwest Europe is close to 50%, Southern Europe quite a bit more, Baltic Europe less. So, yes, these calculators, even when they're not wrong, can be very easily misinterpreted.

As to ancient samples, yes, they help, but I'm certainly not convinced that people know how to properly create calculators using them, nor that people know how to interpret them. The DNA Land "analysis" of ancient samples is just the latest example. Faced with an ancient genome, the algorithm does the best it can, and assigns to groups with a lot of that ancestry, but faced with all the formal testing that's been done, is someone really going to claim that modern Slavs are 100% ancient Yamnaya or Corded Ware?

A little knowledge can be a dangerous thing.

It's totally cool, I wanted to figure out weather we were on the same page. I guess I was a little off keister. Good to know that "the more the better" plan is a myth. As far as figuring out how much German or French you have, I'd check with genetic cousins and/or following your family tree via paper.


Yes the 100% ancient Yamnaya/Corded Ware dilemma sounds like a sticky and dangerous problem. I'm wondering if Gedmatch is willing to rename the samples to make them more user friendly with links to explain each sample; to avoid misinterpretation. Some users like to learn via video and articles.


As far as Corded ware is concerned, I believe Maciamo made a rough map on it a couple of years ago; interpreted as r1a Corded Ware/North Yamna tribes specifically minus the Mesolithic Europeans and Farmers that got absorbed. http://www.eupedia.com/europe/autosomal_maps_dodecad.shtml#East_European

Anyways, hope this helps. Good luck in fixing the interpretation problem. ^_^

davef
29-08-16, 15:50
What makes this even more confusing is that different populations can have similar genes so it you're Russian and the test only uses Polish samples, it'll guess you as Polish.

Also i wouldn't be too optimistic in the event that Ancient Greek bones are dug up and you share in certain areas with a dead Ancient Greek woman from Athens that are "distinguishably" greek bc it could just be due to having significant mediterranean ancestry that she also has, not due to being in with Plato.

And yes obviously sharing with a modern population that lives in the same area as the (insert famous civilization here) doesn't tell as much as far as being related to those ancient people is concerned. That civilization could've experienced drastic changes in its makeup over time. Its a simple concept, don't know why people don't understand it.

Twilight
29-08-16, 18:18
What makes this even more confusing is that different populations can have similar genes so it you're Russian and the test only uses Polish samples, it'll guess you as Polish.

Also i wouldn't be too optimistic in the event that Ancient Greek bones are dug up and you share in certain areas with a dead Ancient Greek woman from Athens that are "distinguishably" greek bc it could just be due to having significant mediterranean ancestry that she also has, not due to being in with Plato.

And yes obviously sharing with a modern population that lives in the same area as the (insert famous civilization here) doesn't tell as much as far as being related to those ancient people is concerned. That civilization could've experienced drastic changes in its makeup over time. Its a simple concept, don't know why people don't understand it.

Idk, although I see where you are coming from And I realize that it's not the most ideal. However, with DNA testing, we have opened up a instinctual void; the desire to know where you came from as shown in mythology. Unfortunately there is really no turning back, unless if DNA testing gets banned. As far as Mediteranian in concerned, I agree that's a broad gene however since the Romans dominated the Mediteranian including the natives; Greek, Illyrian, Itallic tribes, Iberians, etc. could be any one of those tribes and not nessesarily Greek. Ultimately if you really wanna know where you roughly came from, you need to look at the history of your ancestrial land.

davef
29-08-16, 19:21
Yeah and its certainly better than nothing! I think it's the broad stuff that sometimes confuses these algorithms like how it detected "Balkan" ancestry in angela. The Balkan is probably a general mix that is found in both north italy and the Balkans but in the db it's found more frequently in albanians, Bulgarians etc. More north italian samples should eventually make it hide it under "italian".

Alan
29-08-16, 19:25
Didn't we had all this sorted out already. But somehow some people don't seem or want to understand that. Calculators based on modern populations, be they of companies or bloggers are of absolutely no use for ancient samples, especially if they are centered around one part of the world.

DNA land as well 23andme are obviously designed for Europeans and Ashkenazis. So in other words what the DNAland algorythm will try to do is FIRST explain the ancestry of an ancient samples with modern populations of Europe and afterwards if their is still need for extra non European ancestry it will add some of that too. It absolutely doesn't tell you where this components came from, from let's say 1000 AD or older. If we know had a calculator centered around South_Central Asians it would do exactly the same, explain ancient samples FIRST with proxy populations it has collected from South_Central Asia and every additional non South_Central Asian DNA with ancestry from "Outside".

That is how it works. 23andme on the other hand is even more NW European and Ashkenazi centered because most of their costumers have ancestry from them. So it simply tries to explain FIRST your ancestry with NW European or Ashkenazi proxies and everything that doesn't fit into this with additional outside admixture. THis is why North Italians are shown as 25% "NW EUropean mixed" or having Ashkenazi ancestry. This is why Greeks are shown as 40% Near Eastern, because they have additional Near Eastern admixture that isn't found in NW Europe or the Ashkenazis.

Using calculators made for modern populations, for ancient samples, is absolutely bogus idea to begin with.

Angela
29-08-16, 23:52
Actually, I think what 23andme does can be visualized like a giant pca. Imagine all the samples of people with all four grandparents from one place as small coins placed on a pca in relation to one another. Imagine that you can see them on the Italian peninsula and Sicily. Since most of their samples are from their own testees, many of their samples are actually from southern Italy, although they have over 100 Tuscan academic samples. So, the "centrum" for the samples is, I think, somewhere around Lazio and somewhat south of that. They label that "Italian". The samples to the north will have less of that and the closest matches for the rest of their ancestry is all the North Western European samples they have, plus Balkan, plus Southwestern ancestry. People from the Veneto often have Balkan as their highest "secondary" ancestry, etc. I have all three of those showing up, fwiw, but my highest is SW European. The samples to the south of have increasingly more "Caucasus" than does the "Centrum", as well as a percent or two of SSA in certain areas, and maybe 2-3% North African in those same areas. They're all "Italian", in that the vast majority of their ancestors have been on the Italian peninsula for far longer than 500 years. This is partly due to the fact that there's so much genetic variation in Italy. If more people from the Veneto or Piemonte or Emilia and Liguria were tested, the centrum would get bigger, and more of my genome would get labelled "Italian". In a way, while that would be nice, it would do more to obscure the definite Italian cline. The differences are smaller in the British Isles, so at the level of resolution 23andme uses, they can't distinguish between the different areas at all.

I actually think that at least for Italians, 23andme is better than DNALand, and actually better than all of them. The only thing is that there are people who don't understand their algorithm, or who wish to deliberately misinterpret the results. I see that anthrogenica has become sort of an outpost of theapricity in this regard. All of Italian genetic history, or at least southern Italian and Sicilian genetic history, is being interpreted on the basis of perhaps two to three samples maximum from different cities in Sicily. First of all, we don't have any way of knowing if the specific results are actually from the labeled cities in Sicily. Second of all, they might not be representative samples. Third of all, all four of the person's grandparents may not have come from that specific city. Fourth of all, you can't assume that the people of a city like Catania, or Messina, have not changed for the last thousands of years. Messina, for example, suffered a huge earthquake in the very early 20th century, and others before that. Large percentages of the population left, and a big percentage of those people never returned. On the reverse side, many people moved to Messina from other parts of the island and other parts of Italy as it was rebuilt. You have to know something about the history of the region you're discussing before you can speculate in this way, and even then that's what it is...speculation.

The reason I can be more certain about my own ancestry is because my father's family comes from such an isolated region of Italy that Cavalli-Sforza chose it as a place to study genetic drift. It's not as bad as those Bavarian and Tyrolese north east Italian isolates, but it was pretty isolated. There wasn't anything but mule tracks going in until the late 1920s. How much admixture with other places can have taken place? Even today, you have to go to 1400 meters by winding one lane mountain roads to reach it, and even things like phone service are very iffy. As a result, I can only find one "foreigner" in my father's entire line in the last 500 years, and that one man was from the Veneto. My mother's area is not quite as isolated, but almost...Zeri, a rural area outside of Pontremoli, has been the subject of a number of genetic papers. If someone wants to know the genetics of medieval Italians from the whole swathe of territory from Emilia through the Apennines to far Northwest Toscana and eastern Liguria, I'm their gal.

G.Paoli, A.Cecconi, M.T.Del Santo, P.Francalacci, S.M. Borgognini Tarli. (1988) Genetic markers in an isolated population of the Northern Apennines (Zeri, Upper Lunigiana - Italy).

Cavalli-Sfroza:http://press.princeton.edu/titles/7800.html

Unfortunately, from my point of view, he's turned his attention to Sardinia.
http://www.pubpdf.com/pub/23249956/Genetic-characterization-of-northeastern-Italian-population-isolates-in-the-context-of-broader-Europ

You can't compare that to coastal Sicilian cities. You have to leave this kind of speculation to experts, which usually means not only people working in population genetics computation, but natives of that country. In the recent paper looking at Lombard mtDna in Italy, the researchers knew that if they wanted to find traces of a uniparental marker,which is so subject to drift, they wouldn't find it in Torino, where half the population is from the south. They couldn't even track it down in medium sized cities, because there's been too much admixture. They could only find it by looking at the dna of people who have held the same land grants since the Lombards gave them out.
http://www.eupedia.com/forum/threads/32214-Early-Medieval-Migrations-into-Northern-Italy-through-uniparental-markers

The whole amateur field is full of misunderstandings and deliberate misinterpretations like this because so many people have agendas. That's why the finalinterpretations should be left to the academicians.

Fire Haired14
30-08-16, 07:37
@Angela,

What percent Italian do you score? I'm sure 70%+. If so, then it is telling you that you are overwhelmingly Italian. Anyone who scores that high in something probably already knows they're 100% that thing. The 23andme European regions Scandinavian, British/Irish, French/German, Italian, East European, and Iberian do good for White and Latino American customers. I've seen plenty of results and the percentages are all consistent with genealogy. The only bad error in their chosen regions is Eastern Europe. If Americans want detailed info about where their ancestors lived in the last 500 years the paper trail is their only option.

Sile
30-08-16, 12:23
@Angela,

What percent Italian do you score? I'm sure 70%+. If so, then it is telling you that you are overwhelmingly Italian. Anyone who scores that high in something probably already knows they're 100% that thing. The 23andme European regions Scandinavian, British/Irish, French/German, Italian, East European, and Iberian do good for White and Latino American customers. I've seen plenty of results and the percentages are all consistent with genealogy. The only bad error in their chosen regions is Eastern Europe. If Americans want detailed info about where their ancestors lived in the last 500 years the paper trail is their only option.

The noted 500 years only by 23andme is rubbish ...............look at mine and I have registered paper trail of paternal and maternal sides from North-East Italy since 1680 to present

99.6%European

Southern European
21.0%Italian
6.0%Balkan
4.1%Iberian
31.5%Broadly Southern European

Northwestern European
5.4%French & German
3.0%British & Irish
18.2%Broadly Northwestern European

0.6%Eastern European

0.1%Ashkenazi

9.7%Broadly European

0.3%East Asian & Native American
East Asian0.3%
Broadly East Asian

21% Italian ? .............how so, for 300 years out of 500 years.............clearly the maths do not work ................I calculated 23andme goes back 2100 years

Angela
30-08-16, 15:45
@Angela,

What percent Italian do you score? I'm sure 70%+. If so, then it is telling you that you are overwhelmingly Italian. Anyone who scores that high in something probably already knows they're 100% that thing. The 23andme European regions Scandinavian, British/Irish, French/German, Italian, East European, and Iberian do good for White and Latino American customers. I've seen plenty of results and the percentages are all consistent with genealogy. The only bad error in their chosen regions is Eastern Europe. If Americans want detailed info about where their ancestors lived in the last 500 years the paper trail is their only option.

Your certainty is once again misplaced. I'm in the mid-50s, which is an absurdity. Let's just imagine that I'm an adoptee hoping to discover my genetic ancestry. This result would be 50% wrong. Get it?

Full Tuscans south of me get 68%, the Lombards with whom I share 44%, Ligurians get approximately my score, the Emilians and even some Lunigianesi with whom I share about 50%. You only get to 75% Lazio and south.

This is all directly correlated to the fact that they have hundreds of samples for southern Italy, and at least 120 samples for Tuscany. They have extremely few northern Italian samples; they haven't bothered to get even all the academic samples, so they basically have a handful of academic samples and the few northern Italians who have sampled.

Given Italy's genetic variation, this is woefully inadequate. It's especially inadequate for northern Italy as it seems there's more variation there than in the south.

Plus, you're completely missing Pickrell's point. If you know all four of your grandparents are from Germany, or Ireland, or Italy, you don't need these tests. THE BIGGEST MARKET IS ADMIXED AMERICANS, or people mad about genealogy who want to break through walls in their paper trail, or adoptees, or people who have suspicions as to whether or not they're illegitimate. Stop thinking that most of them have any interest in the racist tinged concerns of a lot of people on "anthrofora". They don't give a darn how much hunter-gatherer they have. They want to know, like the man in the ad, if they should buy lederhosen or a kilt.

Most of the un-hyphenated Americans I know who have taken the test are extremely disappointed. They have only the vaguest knowledge of their ancestry: they know there's some English in there, some Ulster Scots maybe, some Welsh names, sometimes Irish and German most recently. They want to know the PERCENTAGES of each...Now, that may not be important to you, but it's important to them, and 23andme can't answer those questions, and neither can any other company.

Did you bother to read Pickrell's article? I'd bet a considerable sum that it's written partly out of frustration with the hundreds, if not thousands, of e-mails he's received at DNALand asking for just that kind of certainty. I know that 23andme boards were full of complaints. He basically says that if you want that kind of information you should do a really good and exhaustive family tree of all your lines. Of course, that's a lot of hard work, especially for Americans who have had such mobile ancestors, and records here don't necessarily say the exact ethnicity. Sometimes you only have surname information. What he doesn't mention is that there are some NPEs in there, although not as many as used to be thought, and German people who might have Anglicized their names, or even African-Americans with enough European "blood" to pass.

Forget about these tests helping people break through "genealogical" walls. In all the years I was on 23andme I can count the success stories on the fingers of one hand. What it does sometimes do is reveal hidden African ancestry, or hidden Ashkenazi ancestry, or, indeed, tell Latinos how much "Indian" they have, versus European, because those alleles "stick out" for one reason or another, and it also sometimes tells people they are not the biological children of their fathers. That often creates havoc for the entire family. I honestly think that in most cases people would be better off not knowing.

If you had spent years on 23andme forums, or were a geneticist like Joe Pickrell, you'd know these things.

davef
30-08-16, 16:19
Just a thought...the italian category seems almost meaningless. Italy is diverse and large. How does one get 100 percent italian on this test? I think you would need to be an italian south of Tuscany but north of southern italy since there's Tuscans in the mix. I doubt a Sicilian or someone from the north would get 100 percent but who knows really?

I just find it crazy to have one label for a big nation, even the north is diverse, Sile got 21 percent italian and angela gets roughy 55 percent (she said mid 50s). It makes sense to have a single category for the british isles bc the irish and the british are strongly related.

Maybe they should have separate categories ie south italy, north italy, etc even that's shaky since Sile isn't all that close to angela at least according to this test though they're both north italian.

I wouldn't take the italian score seriously. Its meaningless.

Angela
30-08-16, 16:38
It all depends how you define northern Italian. If you go by the definitions Italians use, where Tuscany is central Italy, and you take provincial political borders seriously, i.e. you put the Lunigiana into Tuscany even though it only became Tuscany politically about 400 years ago, I'm three quarters North Italian and 1/4 Central Italian.

However, 100% Ligurians also have scores close to mine.

Sile's scores are very low. Even the Friulani score in the mid-thirties to forty. I would think mid-forties for northern Piemontese, Lombards etc. is pretty accurate.

This comes with the usual disclaimer that these are only the scores I've seen and are based on ancestry reporting by the people whose scores I've seen, and so it's not a representative sample from which one can draw vast conclusions.

The point is that there are incredibly few northern Italians in the 23andme reference sample, and so 23andme doesn't include them in the definition of "Italian".

You have to know how these algorithms work to interpret them properly. "Italian", like "German", or "Spanish", is a political term. Genetic clines don't respect modern, national borders. Catalans are different from people from Murcia or Extremadura. People from Alsace are different from people from Provence.

This is what Pickrell was getting at.

Ed. Southern Italians are no more "Italian" than I am. There's just more of them being used as a reference for the genomes of people from the Italian peninsula and Sicily.

Regio X
30-08-16, 18:53
It all depends how you define northern Italian. If you go by the definitions Italians use, where Tuscany is central Italy, and you take provincial political borders seriously, i.e. you put the Lunigiana into Tuscany even though it only became Tuscany politically about 400 years ago, I'm three quarters North Italian and 1/4 Central Italian.

However, 100% Ligurians also have scores close to mine.

Sile's scores are very low. Even the Friulani score in the mid-thirties to forty. I would think mid-forties for northern Piemontese, Lombards etc. is pretty accurate.

This comes with the usual disclaimer that these are only the scores I've seen and are based on ancestry reporting by the people whose scores I've seen, and so it's not a representative sample from which one can draw vast conclusions.

The point is that there are incredibly few northern Italians in the 23andme reference sample, and so 23andme doesn't include them in the definition of "Italian".

You have to know how these algorithms work to interpret them properly. "Italian", like "German", or "Spanish", is a political term. Genetic clines don't respect modern, national borders. Catalans are different from people from Murcia or Extremadura. People from Alsace are different from people from Provence.

This is what Pickrell was getting at.

Ed. Southern Italians are no more "Italian" than I am. There's just more of them being used as a reference for the genomes of people from the Italian peninsula and Sicily.I share with seven, say, "Venetians", and, in fact, the Italian % vary significantly even within a same province, as in your area, likely. I'm not sure, but I would bet people from central/east TV to VE, from VR and RO tend to have more Italian % than Venetians from another areas, and the point is that they are all Italians, obviously.
Rounding off, here they are (considering the grandparents):
- down 33% Italian (b. in North TV);
- up 47% (child of the individual above; a parent from North TV, other from East TV) - the highest % (this individual has just a little, little more than 20% of NW Euro);
- down 26% (1/2 Central BL S. Alps, 1/2 VE) - Edited;
- down 33% (as per surname and overall result, Venetian);
- down 13% (both parents from N. TV) - the lowest (this individual has a bit more than 50% NW Euro), even lower than Sile's;
- down 29% (Venetian);
- up 30% (1/2 N. TV, 1/4 Central TV, 1/4 UD plain core).

And:
- down 33% (my father; 1/4 E. TV, 1/4 N. MN, 1/2 N. TV);
- up 32% (my mother; 1/4 N. PD, 1/4 W. TV, 1/4 N. TV, 1/4 Central BL S. Alps);
- 38% (mine).

Variations (% - speculative level), rounding off:
- S. Euro: 40-70;
- NW. Euro: 20-50;
- E. Euro (5 individuals): 0-2;
- Ashkenazi (1): 0-2;
- North African (1): 0-0.3;
- East African (2): 0-0.2;
- East Asian (1): 0-0.1;
- Native American (1): 0-<0.1.

If there were more reference samples from Northeast Italy, the Italian % would grow.

Sile
30-08-16, 20:21
I share with seven, say, "Venetians", and, in fact, the Italian % vary significantly even within a same province, as in your area, likely. I'm not sure, but I would bet people from central/east TV to VE, VR and RO tend to have more Italian % than Venetians from another areas, and the point is that they are all Italians, obviously.
Rounding off, here they are (considering the grandparents):
- down 33% Italian (b. in North TV);
- up 47% (child of the individual above; a parent from North TV, other from East TV) - the highest % (this individual has just a little, little more than 20% of NW Euro);
- down 26% (1/4 BL valley, 1/4 BL S. mountains, 1/2 VE);
- down 33% (as per surname and overall result, Venetian);
- down 13% (both parents from N. TV) - the lowest (this individual has a bit more than 50% NW Euro), even lower than Sile's;
- down 29% (Venetian);
- up 30% (1/2 N. TV, 1/4 Central TV, 1/4 UD plain core).

And:
- down 33% (my father; 1/4 E. TV, 1/4 N. MN, 1/2 N. TV);
- up 32% (my mother; 1/4 N. PD, 1/4 W. TV, 1/4 N. TV, 1/4 BL S. mountains);
- 38% (mine).

Variations (% - speculative level), rounding off:
- S. Euro: 40-70;
- NW. Euro: 20-50;
- E. Euro (5 individuals): 0-2;
- Ashkenazi (1): 0-2;
- North African (1): 0-0.3;
- East African (2): 0-0.2;
- East Asian (1): 0-0.1;
- Native American (1): 0-<0.1.

If there were more reference samples from Northeast Italy, the Italian % would grow.


My line is in majority north of treviso
, with my father at 26% and one of my sons at 33%


Map below is my lineage
red circle = my paternal line for over 300 years .....62 is the main area presently , but the oldest is the 6 NW of 62

blue circle = my maternal side for over 300 years ........13 is the oldest and 24 is the present

the non circled is my wife's paternal and maternal
31 = paternal origins

13 in the south = maternal origins

30 = present line


http://i103.photobucket.com/albums/m153/vicpret/pedig_zps2mng4cks.jpg


Note:

I did not add Siror and Belluno further north of this map for my paternal line
(http://s103.photobucket.com/user/vicpret/media/pedig_zps2mng4cks.jpg.html)

Regio X
30-08-16, 21:52
My line is in majority north of treviso
, with my father at 26% and one of my sons at 33%


Map below is my lineage
red circle = my paternal line for over 300 years .....62 is the main area presently , but the oldest is the 6 NW of 62

blue circle = my maternal side for over 300 years ........13 is the oldest and 24 is the present

the non circled is my wife's paternal and maternal
31 = paternal origins

13 in the south = maternal origins

30 = present line


http://i103.photobucket.com/albums/m153/vicpret/pedig_zps2mng4cks.jpg


Note:

I did not add Siror and Belluno further north of this map for my paternal line
(http://s103.photobucket.com/user/vicpret/media/pedig_zps2mng4cks.jpg.html)Nice map!

Yes, people - especially from the Po valley or near it - moved around. Also my family, of course. The locations of the most distant known ancestors in male and female lines of both my parents, for example, don't correspond to the locations of their paternal grandfather and maternal grandmother.

How much Italian % your wife got?

PS: I edited a "proportion" in my previous post. Sorry for the mistake.

Zionas
29-10-16, 23:37
​There's been even less studies focused on Asians.