Wednesday, November 16, 2016

Ancient admixtures look shifty

It is hard to believe in some ancestry results.   FamilyTreeDna's new Ancient Origins give me following results

Metal Age Invader 12%
Farmer 30%
Hunter-Gatherer 54%
Non-European 4%

Regarding Metal Age Invaders they refer to the Metal Age Yamnaya culture, regarding Farmers to the Neolithic Anatolian migration to Europe and regarding Hunter-Gatherers to ancient LaBrana, Loschbour and Motala samples.   Regarding non-European proportion they give a hint to look at myOrigins, which is FamilyTreeDna's admixture analysis based on present-day populations.  My myOrigins give me only one non-European group, Middle Easterners.  I doubt it, the non-European in my Ancient Origins test is likely Asian.

Going further in analyzing results I compared my Ancient Origin results to  scientific papers,  Haak et al. 2015 giving comparable results.  Haak et al.  gives following results for Finns:

EN (Farmers) 31.5%
Nganasan (Asian) 10.2%
WHG (Hunter-Gatherer) 7.9%
Yamnaya (Metal Age Intrurers) 50.4%

Respectively Norwegians get in this study
 
EN (Farmers) 48.2%
Nganasan (Asian) 4.2%
WHG (Hunter-Gatherer) 0%
Yamnaya (Metal Age Intrurers) 47.5%

We can see a huge transition between Yamnayas/Iron Age Intruders and Hunter-Gatherers between Ancient Origins and Haak et al.  I know something about the method used by Haak et al., but I have no idea what FamilyTreeDna did. However, if I try to guess, I would say that they could have used a very drastic LD-pruning.  I can get similar differences by heavily pruned data and it makes sense.  Metal-Age invasion to Europe happened during the Bronze Age, thousands years later than the arrival of hunter-gatherers.  So it is reasonable to assume that we have still much more Bronze Age genetic drift than drift from hunter-gatherers, thus removing LD removes more ancestry of Metal Age Intrurers.  Pruning present-day samples does't have same effect due to more similar genetic composition.

I made also some admixture tests.   Pruning LD gives a big change in ancient admixtures.

My result without pruning

Anatolian_Neolithic 31.4
BA_East_European_Steppe 44,8
East_and_Southeast_Asian 10,8
Western_Hunter_Gathrerer 13

and after pruning


Anatolian_Neolithic 27.5
BA_East_European_Steppe 25.9
East_and_Southeast_Asian 7.8
Western_Hunter_Gathrerer  38.8

I am not saying that the difference between results of FamilyTreeDna and Haak et al. is caused by pruning, because I don't know it.  I only state that pruning ancient samples is risky.




Wednesday, November 9, 2016

Project admix results, revised

My previous test was missing of German reference samples.  Together with the fact that my Swedish reference samples seem to be somewhat off, this gave results biased towards Balto-Slavs.  I have now added German samples available from Pagani et al. 2016 and have rerun all project samples, plus two new Finnish samples. Additionally I tested three Finnish samples introduced by aforementioned study.  Soon after downloading those samples I understood that they don't represent average Finns.  So this point is included after project results.


I had difficulties in editing columns and after some useless efforts I copy-pasted all in plain text format.


A new grouping, Karelian-Finnic indicates a sum of Karelian and Veps people.



FI1
    
Finland     57,0
AMBIG_Europe     25,0
Balto-Slavic     12,9
AMBIGUOUS     2,6
Baltic-Finnic     2,5

   
FI2
    
Finland     37,2
AMBIG_Europe     28,0
Balto-Slavic     14,8
NW-Atlantic-Europe     10,6
AMBIGUOUS     5,6
Saami     3,9

   
F3

Finland     62,3
AMBIG_Europe     33,0
AMBIGUOUS     2,4
Baltic-Finnic     2,3

   
F4
    
Finland     47,2
AMBIG_Europe     18,9
NW-Atlantic-Europe     18,1
Northeast-Europe     15,8

   
FI5
  
Finland     53,8
AMBIG_Europe     33,1
Baltic-Finnic     11,7
AMBIGUOUS     1,4

   
FI6
  
Finland     43,0
AMBIG_Europe     36,0
Baltic-Finnic     12,5
NW-Atlantic-Europe     7,9

   
FI7

Finland     78,7
AMBIG_Europe     17,4
TunNenets     3,4

   
FI8

Finland     56,5
Karelia     25,4
AMBIG_Europe     17,4

   
FI9
    
Finland     42,1
AMBIG_Europe     27,7
Karelia     24,5
Karelian-Finnic     5,0

   
FI10
  
Finland     43,1
Saami     21,5
AMBIG_Europe     10,9
Karelian-Finnic     10,2
AMBIGUOUS     10.0
AMBIG_Siberian     4,3

   
FI11
 
Finland     63,7
AMBIG_Europe     31,7
AMBIGUOUS     2,8
Baltic-Finnic     1,8

   
FI12
  
Finland     71,6
AMBIG_Europe     18,0
Central-Europe     10,2

   
FI14

Finland     69,8
Balto-Slavic     16,0
AMBIG_Europe     11,3
Baltic-Finnic     1,6
AMBIGUOUS     1,3

   
FI15

Finland     62,0
Karelian-Finnic     21,2
AMBIG_Europe     14,9
AMBIGUOUS     1,9

   
FI16

Finland     43,1
AMBIG_Europe     22,9
Estonia     21,8
Karelia     10,3
AMBIGUOUS     1,8

   
FI17
 
Finland     33,9
Central-Europe     24,0
Karelia     13,8
Baltic-Finnic     9,8
AMBIG_Europe     9,5
RU_Pinega     5,6
AMBIGUOUS     2,1
Karelian-Finnic     1,3

   
FI18

Finland     46,1
Karelian-Finnic     19,7
Balto-Slavic     14,5
AMBIG_Europe     8,8
Baltic-Finnic     6,5
Saami     3,7


FI19

Finland    0,62
AMBIG_Europe    0,20
Northeast-Europe    0,08
RU_Pinega    0,05
Saami    0,03
AMBIGUOUS    0,02

   
FI20
 
Finland     57,8
AMBIG_Europe     21,8
Balto-Slavic     10,9
AMBIGUOUS     5,1
Baltic-Finnic     4,3

   
FI21
 
Finland     53,1
Karelia     28,0
AMBIG_Europe     10,7
Northeast-Europe     4,8
AMBIGUOUS     2,2
Karelian-Finnic     1,2

   
SC2
 
NW-Atlantic-Europe     32,8
Central-Europe     32,5
Balto-Slavic     19,3
AMBIG_Europe     13,3
AMBIGUOUS     2,1

   
SC3

Baltic-Finnic     27,6
Central-Europe     21,2
AMBIG_Europe     19,3
Norway     17,5
NW-Atlantic-Europe     12,9
AMBIGUOUS     1,6

   
SC4

Norway     53,0
Central-Europe     18,3
Balto-Slavic     13,7
NW-Atlantic-Europe     8,1
AMBIG_Europe     6,5

   
SC5
  
AMBIG_Europe     28,9
NW-Atlantic-Europe     18,3
Central-Europe     18,3
Ireland     14,1
GermanyAustria     11,5
Northeast-Europe     7,9
AMBIGUOUS     1,0

   
SC6
  
Central-Europe     31,5
NW-Atlantic-Europe     24,7
AMBIG_Europe     16,5
Finland     14,5
Balto-Slavic     11,9
AMBIGUOUS     1,0

   
SC7
  
AMBIG_Europe     29,7
NW-Atlantic-Europe     26,1
Sweden     20,5
Orcadian     11,0
Central-Europe     10,7
AMBIGUOUS     1,9




Additionally some freely available genomes, only for checking the method.

   
Genomes Unzipped, VXP
    
North-Italy     24,9
Central-Europe     20,7
AMBIG_Europe     18,4
Norway     13,7
NW-Atlantic-Europe     12,0
South-Europe     6,6
AMBIGUOUS     3,8

   
Genomes Unzipped, JKP
    
Central-Europe     28,9
South-Europe     19,8
NW-Atlantic-Europe     19,1
Spain     12,5
AMBIG_Europe     11,3
AMBIG_WEURASIA     5,8
AMBIG_SEURASIA     2,0                                      


Razib Khan, downloaded here.
    
Indian     35,6
Sindhi     22,3
Cambodian     12,8
AMBIGUOUS     10,6
Burusho     8,6
IndianJew     6,3
AMBIG_Southeast-Asian     2,4
AMBIG_EEURASIA     1,4


Blaine Bettinger, downloaded here.         
He looks British, with a small portion of Native American.
    
Central-Europe     24,9
Kent     24,1
AMBIG_Europe     21,2
Welsh     9,3
AMBIGUOUS     7,9
Ireland     7,3
Atlantic-Europe     3,3
Native-American     1,9



Tests using Pagani et al. Finns as a Finnish reference   
   
Pagani-FI12
 
Karelia    28,0
AMBIG_Europe    23,8
Central-Europe    17,8
Baltic-Finnic    12,6
Finland    12,1
Karelian-Finnic    3,4
AMBIGUOUS    2,4

   
Pagani-FI14
  
Estonia    23,7
AMBIG_Europe    22,5
Karelia    18,6
Central-Europe    18,5
Finland    7,9
Karelian-Finnic    4,7
AMBIGUOUS    4,1

   
Pagani-FI21
  
Karelia    46,3
AMBIG_Europe    16,1
Finland    10,4
Baltic-Finnic    8,7
Northeast-Europe    8,5
Saami    4,3
AMBIGUOUS    2,9
Karelian-Finnic    2,8

I tested three Finns, seen above, two of them typical Western Finns without any obvious foreign admixture and one should be a typical Finn from East Finland. The first row below shows the average result using average Finnish reference picked from 1000-genomes and the second row shows the average result after changing the reference to Finnish samples of Pagani et al.
   
FI12, FI14 and FI21, average Finnish result when using average Finnsh reference    64,8

FI12, FI14 and FI21, average Finnish result when using Pagani Finnish samples as a reference    10,1

In this particular case, while Pagani Finns almost fully mismatch with average Finns, it also eliminates Finnish admixture of Swedish results where it is present in analyses based on average Finnish reference, in some cases substituting Finnish admixture by Karelian and Veps.  This is really odd.


A map giving an estimate of admixture regions in Europe