Ancient DNA of Father Abraham

[Daniel72]

With PF and the like, your data goes behind a curtain, the wizard pulls a few levers, and out rolls his hypothesis. What the wizard does behind the curtain remains a magical mystery and cannot be questioned.

Well, theoretically everyone could download a software like admixture and do his own conclusions with it (if he gets some participants for reference....)

The thing is....

Humans can compare 67 Y-DNA markers or a handfull of SNP.

But "Behind the Curtain" a computer takes 10 hours of calculation power to compare several hundred thousands of alleles to that of other participants.
There is no human on earth who could do this manualy.

EDIT: Just realised this tread seems to concentrate on Jews.
Well, Jews is a special case anyways, because its a people that dont all live in a single country.

[gtc]

Well, theoretically everyone could download a software like admixture and do his own conclusions with it (if he gets some participants for reference....)

And how many are needed (from as many global regions as possible) to form any sort of reasonable conclusion?

But "Behind the Curtain" a computer takes 10 hours of calculation power to compare several hundred thousands of alleles to that of other participants. There is no human on earth who could do this manualy.

Yes, but more importantly how does the wizard audit the results? What is the standard reference data for calibration of his algorithms? As far as I know there isn't any. It's all best guess stuff, so I take it with a grain of salt.

EDIT: Just realised this tread seems to concentrate on Jews.
Well, Jews is a special case anyways, because its a people that dont all live in a single country.

I agree. It was the sudden appearance of x% Jewish in the Illunima PF results that prompted the many questions. I don't think FTDNA did itself any favors by using that terminology. As has been said in another Forum, they (and others) should probably stick to geographical regions with this stuff.

[Daniel72]

And how many are needed (from as many global regions as possible) to form any sort of reasonable conclusion?

So far it seems to give reasonable conclusions with only 5 people per country as reference.

I would like to see 5 people per local region inside countries for better resolution however. Unfortunately it seems to be close to impossible to get participants with all 4 GP from one region (at least the Germans are all mixed from totaly different parts of Germany)

Yes, but more importantly how does the wizard audit the results? What is the standard reference data for calibration of his algorithms? As far as I know there isn't any. It's all best guess stuff, so I take it with a grain of salt.

All science allways should be taken with a bit of salt. Whatever claims to be the unchangable absolut truth can be called "religion".

Of course its like this, that the calibration is so, that it matches the knowlegde of history and prehistory. Means, if the wizard doesnt agree to the common knowledge, it is calibrated until it says: yes, everything that we knew so far is correct. (its usualy tried that these things prove that the old groups like Celts, GErmanics, Slavs, Balts etc do exist on a genetic level or that language groups largely correlate with genetics etc.)

Thats a little danger. Because we already chosed a truth and now only want it to be proven and if it does not prove OUR thruth, there is something wrong with the algorythm and it needs tweaking.

[gtc]

Thats a little danger. Because we already chosed a truth and now only want it to be proven and if it does not prove OUR thruth, there is something wrong with the algorythm and it needs tweaking.

I have seen that in threads on DNA-Forums with settings of k=this, and k=that, and k=something-else and the results being different each time to the point of confusion (mine at least).

[nathanm]

The Independents do not use the same Data set as FTDNA, That's the point.

The independant use their participants as references. Wich is, why most of these projects limits their participants to people who have all 4 Grandparents from one country (such people increase the reference set and improve the overall quality of the project) while mixed people can not be used as a reference and possibly even have a negative influence on the results of others. they for sure have an negative impact on the calculation durations.

So, the indepentant DONT all use the same references, except they all have the same people as participants. Because the participants are meanwhile the core of the references they use.

Actually, most of the independent projects do use the same reference dataset as FTDNA, plus other publicly available academic datasets. The only difference are their participants, which tend to be confined to the specific geographic area the project is focused on. And yes, you too could download much of the reference data and run Admixture on your own. That's exactly what I've been doing (which I really need to write up soon). Dodecad has put together the most extensive set of new populations, because it has the most participants. You can see all the populations it uses in this spreadsheet.

[Javelin]

Whatever reference data FTDNA is using, and/or whatever parameter settings they have chosen to use for their matching algorithms, as per MMaddi's post they have suddenly created Jewish associations for a lot of people who simply can't swallow that conclusion.

Leaving aside Italians, most people here who show Near East will say something like:

90% European (Orkneys)
10% Middle East (Adyghei, Druze, Iranian, Jewish, Palestinian)

This averages out to a point on a map 10% of the way from the Orkneys to the Near East/Caucasus, i.e. somewhere near England or northern France, which is the area most colonial Americans, who are most of the test takers, average out to.

Human genetic overlap is continuous. Points of overlap between neighboring nationalities are usually about as great as the points of non-overlap. This means there are different forumlas you can use to reach the same map points, each with a different probability. Population Finder displays one of many possible formulas using its algorithm's highest probability.

It's not always right, but the averaged spot on the map is usually very close, even for people with complex ancestry.

If the thought of being Orkneys + Middle Eastern makes someone uncomfortable, they can take comfort in knowing there are probably also formulas to express the same person as Orkneys + East African + East Asian, with no Middle Eastern at all.

[vinnie]

Leaving aside Italians, most people here who show Near East will say something like:

90% European (Orkneys)
10% Middle East (Adyghei, Druze, Iranian, Jewish, Palestinian)

This averages out to a point on a map 10% of the way from the Orkneys to the Near East/Caucasus, i.e. somewhere near England or northern France, which is the area most colonial Americans, who are most of the test takers, average out to.

Human genetic overlap is continuous. Points of overlap between neighboring nationalities are usually about as great as the points of non-overlap. This means there are different forumlas you can use to reach the same map points, each with a different probability. Population Finder displays one of many possible formulas using its algorithm's highest probability.

It's not always right, but the averaged spot on the map is usually very close, even for people with complex ancestry.

If the thought of being Orkneys + Middle Eastern makes someone uncomfortable, they can take comfort in knowing there are probably also formulas to express the same person as Orkneys + East African + East Asian, with no Middle Eastern at all.

I hear what you're saying, but if PF is going to tell someone that they're of 10% ME origin, FTDNA should have it right, with the caveat that this is all a work in progress. But I've also noticed that most people don't include their margin of error in their posts, which is important to take into consideration.

BTW, there's a group of Scottish Camerons, who are supposedly a noble line, who've turned out to be J1c3d (associated with Jewish and Arabic origins) and not R1b or some other expected Celtic or North Germanic group. Additionally, as Thomas Jefferson was J2, it's entirely possible for people with family origins in the British Isles to have some ME ancestry. One of my own closest matches, sharing the same subclade, is of British origin.

[Javelin]

I hear what you're saying, but if PF is going to tell someone that they're of 10% ME origin, FTDNA should have it right, with the caveat that this is all a work in progress.

I do agree with this. Without wanting to get too technical, if you run phylogenetic trees, Europe looks like a subcategory of the Near East, so it is hard to pull the Mideast out of Europe entirely. It might be helpful to users if they were able to pick from a number of options (such as about 100% English as opposed to 90 Orkneys 10 Middle East) -- or it might be more confusing, and people often identify themselves in ways that oversimplify their true genetic origins.

[Daniel72]

There is a lot of weakpoints in the current way to interpret aDNA.

If one looks at dodecad Oracle results one sees a lot of ways to interpret "German" for example.

German can be decribed as (just some examples):

68.8% Dutch + 31.2% Polish
84.4% Orcadian + 15.6% Russian
65% Scandinavian + 35% Southeast European

And now watch up...

96% Orcadian + 4% Egyptian
97% Orcadian + 3% Ethiopian

Well, even
97% Orcadian + 3% Indian Brahmin

All this is, kind of correct, but the reason is not "recent" admixture but bases on admixture that is centuries or even thousands of years old.

Its just very rough to explain a German as "Orcadian + X". The "X" will be very exotic, without beeing any more recent as 8000 years ago.

And if one thinks, PF does not work like Dodecad Oracle and this kind of stuff, then look here....

I digg this here up from MY Dodecad Oracle results:

6% Bedouin + 94% Orcadian

and thats my PF:
Orcadian 92%, Middle Eastern 8%

Pretty close. Close enough to claim, the general technology is similiar.

[Javelin]

But which one of those ways is most likely? That's the hard part. Probably most of their customers are "Europe (Orkneys)" crowd (i.e. typical white Americans), and the second largest group is "Middle East (Jewish)". So you can see why these are used, i.e. statistical likelihood.

The weakness in any such analysis as you describe is that it is impossible to identify the age of these mixtures with single SNPs. Even blocks of 1-5 cM are likely to be identical by state.

If Population Finder also looked at the same longer identical-by-descent blocks we use to find cousins, and was somehow able to anchor these blocks to countries or regions (perhaps with user input, the way we can do with degress of relation), it could be much more useful, and get beyond many of the problems of proxy populations.

[nathanm]

The weakness in any such analysis as you describe is that it is impossible to identify the age of these mixtures with single SNPs. Even blocks of 1-5 cM are likely to be identical by state.

If Population Finder also looked at the same longer identical-by-descent blocks we use to find cousins, and was somehow able to anchor these blocks to countries or regions (perhaps with user input, the way we can do with degress of relation), it could be much more useful, and get beyond many of the problems of proxy populations.

That would be great if it was possible, but I think it's several steps removed from where we are with current science and technology. In the last issue of the Journal of Genetic Genealogy (JoGG), there was an article called Phasing the Chromosomes of a Family Group When One Parent is Missing. Even with the best technology on the market, you either need to test yourself and both parents, or one parent and three children (four is better), just to determine which parent a specific segment was inherited from.

Going from that knowledge to matching segments with more distant relatives, and inputting accurate information about your common ancestors, is quite a tall order. Paper trails for some ancestors might simply not exist. Plus, the size and complexity of such a database, cost of testing so many people, and reluctance of some to genetic testing will not be inconsequential barriers. Someday, once the $1000 genome is a reality, science and technology have advanced far beyond today's infancy, and genetic testing is far more common, it might be feasible to attempt something like you describe. But it's a long way off in the future.

[Javelin]

This doesn't need phasing: it's a simple IT change. Allow customers to make blocks clickable, and choose the country/population from a pulldown menu.

[nathanm]

This doesn't need phasing: it's a simple IT change. Allow customers to make blocks clickable, and choose the country/population from a pulldown menu.

But how would you know which block came from which parent, grandparent, etc? Unless you have matches for each block, and can both trace your descent from a common ancestor, it wouldn't do any good. The proportion of people who could reliably assign blocks to a specific ancestor, country, or even region would be a tiny fraction of FTDNA's pro-active customer base, to say nothing of the people who've tested and seemingly do nothing with it.

[Javelin]

But how would you know which block came from which parent, grandparent, etc?

Many possibilities:

(1) You have tested one or both parents, or other relatives.
(2) You have succesfully identified the common ancestor.
(3) You have sponsored a known relative.
etc.

[nathanm]

Many possibilities:

(1) You have tested one or both parents, or other relatives.
(2) You have succesfully identified the common ancestor.
(3) You have sponsored a known relative.
etc.

Right, but again, the dataset who'd meet these criteria would be ridiculously small. It definitely wouldn't justify the time and expense to design and develop a new system to track these blocks. However, nothing's stopping a third party from creating such a system. And when you said "it's a simple IT change"; that's the understatement of the month.