DIYDodecad 2.0

[GayeSherman]

DodecadOracle Irish_D mixed mode

What does this mean? Does this assume Irish_D and figure out the rest or does it remove Irish_D from the calculation?

How is it done? I know there's a way to omit certain populations by removing them from the population list. Is there an easier way?

Thanks!

Gaye

[nathanm]

Personally, I would only do the target analysis for larger segments - at least 10cM.

I just compared my brother's byseg run line by line to the Dodecad populations. Even though he's 100% Ashkenazi, there were few individual segments that "matched" an Ashenazi reference population. I'm wondering if I should run a byseg of something greater than 500/50 and see if that smoothes it out.

Several people on DNA Forums ran it at 200/20 and seemed to like the results. One of these nights, I'm going to set it up and let it run overnight. It'll probably take several hours on my 2008 Macbook. But I keep falling asleep before I get a chance.

[nathanm]

I'm not German (well one paternal g-grandparent was 1/2 German...so I'd be 6.5% German??)), but rather English and Czech for my "Western European" parts (but Germany is in-between so I'm assuming that's why that is popping up).

Probably, the way the combination of your parents' DNA looks to the algorithm can be pretty strange. The closest population to my dad is Slovenian, but he has no Balkan ancestors I know of, or even anyone from that general area. He's got three grandparents whose ancestors came from Prussia, all mixed Polish-German, and one whose ancestors came from Stuttgart and Bavaria in Germany. If I sort the population spreadsheet by the various components, it makes more sense. My dad has less East European than Polish_D, but more than German_D; less West European than German_D, but more than Polish_D. So he falls right in between both of his actual ancestral ethnic groups.

Does Dr. McDonald have Czech population sets?? I'm assuming no based on my results and how he said "Czech comes out" in an English and Armenian mix. Czech and German are VERY close, and I noticed there's no Czech populations listed in Dodecad either, so perhaps the German is reading the Czech DNA.

I don't know what populations he is using, just that certain ethnic groups aren't represented well or at all in publicly available, academic datasets. Czech is one of those.

Dodecad does have a "Mixed_Slav_D" population, which might include Czechs, and a "Mixed_Germanic_D" population, which I'm guessing could have Austrians, German-speaking Swiss, etc. Unfortunately, he doesn't say which Dodecad participants comprise the reference populations.

The good news is even though submission to Dodecad is officially closed, Dienekes just wrote a new post today calling for people from specific backgrounds where he almost has enough to form a new reference population. But Czech isn't one of them. If you could convince a few people with full-Czech backgrounds to test and submit their data to the Dodecad project, we would have a "Czech_D" population for comparison.

[nathanm]

What does this mean? Does this assume Irish_D and figure out the rest or does it remove Irish_D from the calculation?

That's showing which other populations are closest to "Irish_D." It can be an interesting exercise for exploring the dataset, and seeing what some of the more obscure populations are.

How is it done? I know there's a way to omit certain populations by removing them from the population list. Is there an easier way?

You can omit populations. From the blog post about Dodecad Oracle, to omit the "Spaniards" and "IBS" (Iberian Spanish) populations from an Oracle run use the following command:

X <- X[setdiff(1:227,which(X[,1]=="IBS" | X[,1]=="Spaniards")),]

You can do the same to omit one or multiple populations, just add the pipe character "|" between them inside the parentheses. This can be useful if most of the entries on your closest populations contain the same population. Omit it, and see what other populations result.

[GayeSherman]

200 by 20?

Why would that give more meaningful results than a higher set of numbers like 1000 by 100?

[nathanm]

200 by 20?

Why would that give more meaningful results than a higher set of numbers like 1000 by 100?

The smaller the window of SNPs, the finer level of detail you'll get. I initially ran it with his suggested values of 500/50. So 1000/100 would only give about half the resolution I already have.

[Taz85]

A small window size may result in more "noisy" results, as it becomes more
difficult to estimate the origin of a segment of DNA of smaller size. On the
other hand, it provides finer-scale resolution for segments of more remote
ancestors. The program does not consider window sizes less than 10 SNPs.



More snp's the better imo. I'm gonna try 2000/200 next.

[Taz85]

lower snps= lower resolution. However It shows remote populations to you with less snps.

[deniseneufeld]

I just read the FAQ about this:

"How difficult it is to tell apart two populations depends on whether or not the populations in question share recent histories (one hundred to two thousand years). When they do, it is much harder to distinguish between them."

So does this also mean that Middle Eastern/ Mozabite could acutally be Jewish? On my family tree I have several ancestors that were Jews from Spain, but my family finder shows 16% Mozabite. My mom's shows 20% Mozabite.

I saw your earlier post on page one again. Had another thought. It is possible to have some Mediterranean show up as Middle Eastern or North African, in fact my Sicilian/Southern Italian does. The area was full of migration, trade, etc., causing a large mix of ancestry that these tests pick up.

I noticed in your signature all the non DIY tests are reading you have Jewish. That could be your Mediterranean being read as such...

So no I wouldn't agree PF is why off, just they way the test differences work.

http://www.familytreedna.com/faq/answers.aspx?id=22#1177

Matt.

[nathanm]

Today on DNA-Forums, the administrator of the Magnus Ducatus Lituaniae Project posted a zip file you can use with DIYDodecad 2.0 to get the percentages for the 7 ancestral components of his project. If you have ancestry in the Baltic region, Ukraine, or Russia, this might be helpful for you. He requires all four grandparents to be from the territory of the former Grand Duchy of Lithuania (1569-1795), so entry into the project wasn't possible for me or my dad. My dad has 3 grandparents from East Prussia, which was partly in the right area, and the paternal line is definitely from there. So this is the next best thing; here are my dad's results:

12.86% Scandinavian
5.09% Volga_Region
3.54% Altaic
23.78% Celto_Germanic
14.45% Caucassian_Anatolian
28.89% Balto_Slavic
11.39% North_Atlantic

And mine:

14.69% Scandinavian
5.27% Volga_Region
4.05% Altaic
28.26% Celto_Germanic
11.17% Caucassian_Anatolian
22.34% Balto_Slavic
14.21% North_Atlantic

If you want to try this out, download the zip file and extract it in your DIYDodecad folder. Then, instead of the command DIYDodecadWin dv3.par, type DIYDodecadWin MDLP.par.

[Taz85]

10.37% Scandinavian
2.76% Volga_Region
1.55% Altaic
27.93% Celto_Germanic
22.61% Caucassian_Anatolian
23.51% Balto_Slavic
11.27% North_Atlantic

[nathanm]

There's yet another BGA project using the DIYDodecad program to calculate its ancestral components. The Eurogenes Genetic Ancestry Project, run by Davidski, has released a calculator for a K=14 Admixture run. He doesn't have the percentages for his reference populations posted yet, but once he does, this may give more finely-grained analysis for people with majority European ancestry. In previous runs of the project, at lower K values, other project members created a spreadsheet to measure the distance from the reference populations, similar to the Dodecad Oracle. Here are my percentages:

36.09% Northeast_Euro
0.00% Sub-Saharan
0.00% Berber
1.36% South_Asian
48.42% West_Euro
0.00% Southeast_Asian
0.01% South_Siberian
0.00% North_Asian
0.33% East_Siberian
0.00% East_Asian
3.54% East_Mediterranean
0.04% Middle_Eastern
7.00% West_Asian
3.22% Volga-Ural

CPU time = 719.45 sec

And my dad's:

41.63% Northeast_Euro
0.24% Sub-Saharan
0.00% Berber
1.62% South_Asian
42.59% West_Euro
0.00% Southeast_Asian
0.11% South_Siberian
0.00% North_Asian
0.25% East_Siberian
0.00% East_Asian
2.50% East_Mediterranean
1.47% Middle_Eastern
7.74% West_Asian
1.84% Volga-Ural

CPU time = 444.05 sec

Note the big difference in CPU time. I guess it takes longer for more admixed individuals like me.

[Taz85]

There's yet another BGA project using the DIYDodecad program to calculate its ancestral components. The Eurogenes Genetic Ancestry Project, run by Davidski, has released a calculator for a K=14 Admixture run. He doesn't have the percentages for his reference populations posted yet, but once he does, this may give more finely-grained analysis for people with majority European ancestry. In previous runs of the project, at lower K values, other project members created a spreadsheet to measure the distance from the reference populations, similar to the Dodecad Oracle. Here are my percentages:

And my dad's:

Note the big difference in CPU time. I guess it takes longer for more admixed individuals like me.

It seems like a pretty good tool. However if your not close in relation to the populations used, it will be pretty useless to some. I dont cluster close to saudis for example, so the population set will not give me an accurate admixture. If the populations were more diverse I think it could be alot more usefull