Starting with admixture analyses

It is the time to go ahead and the next step will be checking the data for upcoming admixture tests between North Europeans, the focus is however on Finns.  For that reason I have collected data from following populations:

Belarussians

Chuvashes

Estonians

Lithuanians

Maris

Mordvas

Finns

Norwegians

Poles
Swedes

Central-West Europeans

I didn’t manage to find German samples which is a pity because Germans operated around 800 years ago widely in Northern Europe (German crusades and Hanseatic traders).   Instead of them I was forced to use UTAH-CEU samples, who are very close Northern Germans on LD-pruned PCA-plots.   A drawback is that they are quite homogeneous and mostly relatives.  I however managed to pick the most diverse group from available UTAH-samples.  It didn’t solve the problem totally, only made it less obtrusive.  

Syrians and Nganassans are used for control purpose.  

Finns are divided into four groups:

•  Eastern Finns being most eastern on global PCA’s

• Western Finnish group 1 represents genealogically selected “purest” Finnish speaking West   Finns.  They live in a wide area, distances between samples are 700 km at maximum

• Western Finnish group2 are the most western Finnish samples on global PCAs

• the rest of Finns are headed Finns, being an intermediate group between Eastern and Western  Finns.

The result shows obvious local genetic drift for UTAH-CEU.  It was expected and would have been much stronger without my prequalification.  I have seen many PCA-plots where UTAH-CEU samples are close Central European, but this happens only in large scale plots where local drift components are diminutive.

Perhaps the most surprising thing is however that Eastern Finns don’t cluster strongly on dimensions 1 and 2, as they usually do in studies.  I expected a strong clustering without  LD-pruning and with full genetic drift.  I have two explanations for this unexpected thing.  The first one is that Eastern Finns have a very distinctive prehistory and it comes into view after the LD-pruning and in this case my data is not pruned.  Secondly the explanation can be that usually  the LD-pruning works poorly with Eastern Finns, because it is always data dependent.  This could be checked by pruning LD using a large amount of East Finns.  

Next I am going to do admixture F3-, D- and Rolloff-statistics to see closer into admixes that standard Admixture analyses can offer regarding local admixtures. 

                                            ********************************

Abbreviations:  BU-Belarussians, CH-Chuvashes, ES-Estonians, LI-Lithuanians, MA-Maris,  MR-Mordva, NR-Norwegians, SE-Swedes, EF-East Finns, FI-Intermediate Finns, PL-Poles, WF-Finns and CEU-Utah-CEU Americans

Click here to see a large picture

Fst-correlation of Admix-analyses

The aim of admix analyses is to find similarities from large data and section it to most probable proportions assuming admixtures. There are likely several ways to see differences, one is overall similarities founding groups and another pure genetic distances. It is likely that the "industrial standard" admix program mostly used today uses both methods, group similarities and Fst-distances. The question is how the result depends on the method. I have run a test using one of the best Gedmatch analysis and compared results between standard results and Fst-corrected results and made euclidean trees to see correlations.  I am not going to reveal this particular Gedmatch admix-analysis, only can say that it is far away being the worst one.   I noticed following differences

- in the stadard output Siberians don't group with East Asians. The corrected output clusters Siberians and East Asians.

- Somalis and Ethiopians group in both outputs, but their common branch groups in the std output with SSA and in the corrected form with North Africans, although in both the distance between the Somali-Etiophian group and the others is moderate.

- in the std output Armenians group with Turks and in the corrected one with Assyrian-Mandean.

- Finns group in the std output with North Russians and after Fst -correction with the group containing Estonians, Balts and North Slavs.

Some other differences were also generated.

This small test probably gives some answers how does the k-grouping  in admix analyses work.  In my opinion it is in practice question about two genetic distances; one between ancestral groups and another inside ancestral groups.



Plain admix:

Fst-corrected:

Ancient Gotlander looks Baltic

I am just working with ancient Scandinavian samples, not finished yet, working with new samples and looking for the best way to analyze them.    Now I'll make only a short  introduction about Ajv58, the ancient Gotlander who lived around 4700 years ago.   After gathering and processing more ancient data I will make a larger analysis.

Ajv58

Click here to see a larger view. 

Comparison to the Skoglund results

      There is some differences caused by different testing arrangement  

•  Skoglund projects ancient samples, I use straight comparison between all samples.  Ajv58 has some impact on present-day samples.
• Skoglund uses huge amounts of Scandinavian, Baltic and Estonian samples.  This has an effect on PCA components magnifying typical genetic components of those populations. 
•            I use only 26kSNPs, while Skoglund obviously use much more SNPs for the background plot.  I have possibility to extent the SNP amount to 160k.  Now I am using the overlapping proportion of my standard set and Ajv58.
  
  
  
  

Two first matters have effect on the distant between present-day and ancient samples and explains the difference between my and Skoglund plots.

Zoomed in picures of mine and Skoglund

IBS

IBS-distances show that Ajv58 is closer North Europeans than LaBrana 1.   He is also more eastern than LaBrana1.  Probably this mirrors Ajv58’s higher ANE affinity or simply the ancestral position of ancient Gotlanders on the Baltic shores.  

Shared ancestries of Swedes, Estonians and western Finns

I am back with the mystery of commercial ancestry tests (while still preparing ancient Scandinavian samples), but I promise that this is the last time if don’t see amazing things happening, like FamilyTreeDna correcting their ideology.  I believe it will never happen.  Comparing two tests I have had possibility to check the 23andme's Ancestry Composition and  it is after the last spring’s revision better than what FtDna now can offer.  FtDna's results are so bad for Finns that I can't even make any reasonable review.  

Let's look at some basics.  There are three things which have a great impact to the quality of genetic ancestry analyses, reference sampling, method and authors understanding about history.  Authors must make compromises regarding reference samples, locations and amount of them and, what is even harder, they have to own some reasonable opinion about the timeline which they are looking for the ancestral idea.  Are they searching 500 or 5000 years back.  The big question is how the reference pack of each ethnic group can be connected to the desired time frame. 

In the case of Finnish and Scandinavian reference samples 23andme has not managed quite well.  My opinion is that in both cases they have found a unique sample set, but not an ancestral one.  Any sample set being unique enough can give unambiguous results, but not right ancestry in a desired time frame or in any time frame.  In the case of Ancestry Composition version two it is however possible to make some tricks to estimate real relations between national or ethnic groups by calculating all overlapping groupings and comparing three or more national/ethnic groups.  It is necessary to use comparison of several ethnic groups to form a big view instead of classifications made straight of reference samples.   This is necessary because the original grouping “Finnish” doesn’t mean Finnish in all cases, for example in Sweden, and in the same way “Scandinavian” doesn’t always mean Scandinavian in Finland and Estonia, for example those two cases.   I am going to do a test by using Scandinavian, Estonian and Western Finnish samples.

Sample set definitions

Scandinavians are a selected group of Ancestry Composition results from Southern Sweden, excluding Scandia, to Western Bothnia (Västerbotten).  They have published genealogical ancestries in their 23andme profiles without known Finnish ancestry.

Estonian group includes all for me now available results, excluding close relatives. There are samples from the north to the southeast and also from the west. 

Western Finnish group includes Finns from Finland Proper, Satakunta and Ostrobothnia without genealogically known Swedish ancestry.

Some basic statistics first.

Swedish samples

	Finnish	Scandinavian	British&Irish-French&German	Broadly North-E	East European	Broadly European	East Asian-Natvive Am.
Average	6,66	58,96	7,03	22,81	2,21	2,54	0,06
Median	7,3	55,8	6,25	22,9	1,45	1,65	0,05
Min	1,3	45,2	0,1	12,3	0	0	0
Max	9,9	75,1	16,3	38,5	7,7	9,8	0,2

Estonian samples

	Finnish	Scandinavian	British&Irish-French&German	Broadly North-E	East European	Broadly European	East Asian-Natvive Am.
Average	21,07	4,94	0,31	8,66	60,78	3,91	0,04
Median	19,9	3,8	0,3	8,6	49,7	3,9	0
Min	2,5	1,3	0	2,2	42,6	1,7	0
Max	43,5	10,4	0,9	16,7	90,6	5,9	0,2

Finnish samples

	Finnish	Scandinavian	British&Irish-French&German	Broadly North-E	East European	Broadly European	East Asian-Natvive Am.
Average	81,18	6,11	0,78	8,53	1,89	1,26	0,03
Median	81,4	5,7	0,5	8,3	1,3	0,7	0
Min	60,7	4,1	0	3,3	0	0	0
Max	90,4	9,8	2,4	15,9	4,9	3,8	0,1

Estonian samples look more heterogeneous than Finnish and Swedish samples, but it is only an illusion caused by the fact that they have not own reference pack.   They are split between Finnish and East European groupings.

Results, shared ancestry between Swedes, Estonians  and Western Finns.

Numbers are calculated by comparing individuals, not by ethnic averages.

Shared between Swedes and Finns:

		Finnish		Scandinavian		British&Irish-French&German			Broadly North-E		East European		Broadly European		East Asian-Natvive Am.
Shared ancestry	6,66		6,11		0,70		8,46	0,99		0,83		0,01		23,75

Total shared ancestry between Swedes and Western Finns is 23,75%

Shared between Estonians and Swedes:

Finnish

Scandinavian

British&Irish-French&German

Broadly North-E

East European

Broadly European

East Asian-Natvive Am.

Shared ancestry

5,99

4,94

0,28

8,58

2,21

1,76

0,02

23,79

Total shared ancestry between Swedes and Estonians is 23,79%

Shared between Estonians and Finns:

Finnish

Scandinavian

British&Irish-French&German

Broadly North-E

East European

Broadly European

East Asian-Natvive Am.

Shared ancestry

21,07

3,82

0,22

6,20

1,89

1,21

0,01

34,40

Total shared ancestry between Estonians and Western Finns is 34,40%

Those shared ancestries don't prove that Swedes, Estonians and Finns have just same named ancestors, only that they come from similar European ethnicities to the shared numbers.

 

There is only a small difference in the shared ancestry between Estonians and Swedes compared to the shared ancestry between Swedes and Western Finns, and both shared ancestries show similar figures for Finnish, Scandinavian and British&Irish-French&German ancestries.  It looks plausible that almost the whole shared ancestry with Swedes is inherited from a deeper history than 23andme assumes.