torstai 21. syyskuuta 2017

Estimates of ancient mixture proportions in present-day Europeans / software

The first Linux script makes possible to generate 23andMe format files from EIGENSTRAT-data and the second one estimates ancient mixture proportions of European people.  Download the full package (without Python interpreter) here.   Instructions are included in the README file.

lauantai 16. syyskuuta 2017

European ancient admixtures

This admixture test was done using the latest data from Reich Laboratory and Dna.Land's Admixture program.  Instead of following the rule and using a mixture triangle Steppe-Farmer-HG I picked also the Central-European Bronze Age to ensure the best coverage of European ancestry.  European Bronze Age was built up from all those three dimensions, so results of all those three roots will be somewhat lower than in tests without the European Bronze Age.  This test works best for Europeans and have obvious blind spots in Africa, Northern America, Near East and Central Asia. I took also some Western Asian samples in the interest of seeing the Iranian and Armenian Neolithic proportions.  

All samples were picked randomly, one sample per population, not as population averages.  No consensus was run.   Both, averaging and consensus calculation would minimize fluctuation in results of minor components, like between Siberian and East-Asian.  The data was converted from the EIGENSTRAT data format I usually use in my tests, of course with exception of the project members.  For that purpose I made a small Bash script which converts Eigenstrat to 23andMe format.  The data and Dna.Land's Admixture program will be soon available here and if anyone see the Eigenstrat-to-23andMe conversion useful I'll include also it to the downloadable library.

sunnuntai 13. elokuuta 2017

Project admixture analyses, revised

Now I used more SNP's with the method coded by Dna.Land authors.  It is now also possible to download all necessary tools for DIY purpose.  It works only on Linux and needs Python to be installed.  Here is a help how to install Python on Ubuntu.

Some comments to understand more about results:

- after a lot of testing I found that the Swedish sample bunch published by the study "No Evidence from Genome-Wide Data of a Khazar Origin for the Ashkenazi Jews" (Behar doesn't fit well with my Swedish project samples and all of them express more Northwest and Central European than the aforementioned Swedish reference.  This happens even if their self-declarations presume some Finnish admixture.   Therefore I decided to label them as East_Scandinavians, which seemed to be correct.  I wonder where they are geographically from.  

- Saami reference samples, unfortunately too few of them were available leading to increased statistic error,  cannot be considered as a source of Siberian.  They represent here a much more diverse source of genetic history.   The small Siberian admixture usually seen in Finnic results is built in Finnish results for the reason that the present-day Siberianness among Finnic people is old and distinct and doesn't match with present Siberians if we simultaneously use also Finnic reference samples.

The summarizing tree:


Finnic 54.6
East_Scandinavian 25.5
Saami 8.0
Northeast_European 5.7
Slavic 2.2
Northwest_European 1.0
Central_European 1.3
AMBIG_European 1.7

Finnic 52.7
Northwest_European 31.3
East_Scandinavian 6.1
Saami 3.8
Northeast_European 2.3
Slavic 1.9
Central_European 1.4

Finnic 72.9
East_Scandinavian 16.3
Saami 3.6
Baltic 3.4
Northwest_European 1.6
Northeast_European 1.7

Finnic 49.1
Northwest_European 25.1
East_Scandinavian 12.9
Northeast_European 5.1
Slavic 2.4
Saami 2.9
Baltic 2.3

Finnic 80.0
East_Scandinavian 13.4
Saami 3.4
Central_European 2.5

Finnic 54.2
East_Scandinavian 27.2
Baltic 9.3
Northwest_European 2.3
Saami 1.8
Northeast_European 1.7
Mediterranean 1.1
Central_European 2.0

Finnic 97.8
AMBIG_European 1.7

Finnic 95.1
East_Scandinavian 2.1
Baltic 1.7
AMBIG_European 1.1

Finnic 85.7
East_Scandinavian 11.9
Baltic 2.1

Finnic 64.0
Saami 31.5
Siberian 2.5
Uralic 1.0

Finnic 92.7
East_Scandinavian 5.2
Saami 2.1

Finnic 83.6
East_Scandinavian 15.2
AMBIG_European 1.1

Finnic 77.7
Baltic 16.8
East_Scandinavian 2.9
AMBIG_European 2.1

Finnic 97.8
AMBIG_European 1.7

Finnic 73.6
East_Scandinavian 14.6
Northwest_European 5.1
Central_European 5.4
Saami 1.0

Finnic 67.8
East_Scandinavian 14.8
Central_European 7.1
Slavic 5.1
Saami 1.4
Mediterranean 1.6
AMBIG_East_European 1.1

Finnic 82.0
East_Scandinavian 12.9
Saami 2.9
Baltic 1.5

Finnic 73.6
East_Scandinavian 14.3
Saami 6.9
Northwest_European 3.8
Slavic 1.0

Finnic 75.8
East_Scandinavian 17.3
Saami 4.6
Slavic 1.8

Finnic 94.0
Saami 2.1
AMBIG_European 2.0
Baltic 1.0

Finnic 94.5
Saami 1.3
Baltic 1.9
AMBIG_European 1.2
AMBIG_East_European 1.1

Finnic 68.8
East_Scandinavian 20.0
Saami 3.6
Northwest_European 3.3
Slavic 2.4
Central_European 1.7

Northwest_European 40.3
East_Scandinavian 22.4
Central_European 15.9
Finnic 9.0
Slavic 4.2
Baltic 3.9
Saami 1.6
Mediterranean 1.7
AMBIG_European 1.0

Northwest_European 52.1
East_Scandinavian 20.5
Finnic 14.8
Slavic 4.1
Central_European 4.3
Baltic 3.9

Northwest_European 59.5
East_Scandinavian 27.3
Central_European 5.4
Baltic 5.8
Saami 1.8

Northwest_European 38.1
East_Scandinavian 32.9
Finnic 11.5
Baltic 9.9
Northeast_European 3.4
Uralic 1.6
Central_European 1.9

Northwest_European 40.8
Finnic 20.7
Northeast_European 13.1
Central_European 9.0
East_Scandinavian 8.4
Slavic 4.0
Saami 2.1
Baltic 1.9

Northwest_European 45.8
East_Scandinavian 31.9
Finnic 11.5
Mediterranean 6.2
Slavic 2.2
Northeast_European 2.1

Although my primary goal was to find out Finnic and Scandinavian admixtures this obviously works fine for almost all Europeans, at least to some extent.

Other samples for a verification purpose:
Irish sample
Northwest_European 90.0
East_Scandinavian 8.7
AMBIG_European 1.3

Western Polish sample
Slavic 49.8
Baltic 18.3
Central_European 14.4
Northwest_European 6.5
Northeast_European 3.5
East_Scandinavian 4.0
Mediterranean 2.3
Uralic 1.1

Sardinian sample
Mediterranean 93.2
Northwest_European 4.5
East_Scandinavian 1.3

Baltic sample
Baltic 70.6
East_Scandinavian 12.3
Slavic 7.9
Northeast_European 6.3
Central_European 2.6
Lithuanian/Yotvingian sample
Baltic 49.0
Slavic 37.5
Central_European 5.8
Mediterranean 4.1
Northeast_European 1.8
AMBIG_European 1.7

Estonian sample
Finnic 41.4
Baltic 19.6
Slavic 16.8
Central_European 9.7
East_Scandinavian 7.8
Saami 2.3
Northeast_European 2.3

Genomes Unzipped sample
Mediterranean 45.7
Northwest_European 19.2
Central_European 19.9
East_Scandinavian 12.3
Slavic 1.9

Genomes Unzipped sample
Mediterranean 37.4
Northwest_European 37.0
East_Scandinavian 15.5
Central_European 9.3

Admixture sums don't give full 100 % because all admixtures below 1% are ignored.

Program downloading and running

Download programs here.  Unzip and locate all programs into a same directory.  To run tests you need use a command line "bash ./ <sample-id>,  where sample-id is the file name holding your genetic data in 23andme format.  The sample file must be compressed with gz file extension (gzip format), but on the command line you give only the sample id (sample-id.gz), not the extension.  The test works fine with following genome builds:  HG18, HG19, GRCh36, GRCh37, but if your genome file is in the FtDna format you have to convert it into the 23andme style.  On Linux it is done easily using four command line entries:

first unzip your genome file and then

cp <original filename> <sample-id>
sed -i 's/\"//g' <sample-id>
sed -i 's/,/\t/g' <sample-id>
gzip <sample-id> 

If your data is already in the 23andme format, but not compressed with gz file extension then you need to unzip it first and run the first and fourth commands explained as above.

edit date 14.8.17 time 17:30

Another Estonian results.  I can only say that it is plausible considering the history

Baltic 37.2
Slavic 29.6
Finnic 22.8
East_Scandinavian 8.1
Saami 1.5

edit 15.8.17 time 17:45

A British results.  It looks like Irish with more Mediterranean and minor Central European admixture..

Northwest_European 81.6
Mediterranean 10.3
Baltic 3.4
Central_European 2.9
AMBIG_European 1.8

tiistai 27. kesäkuuta 2017

Estonian Corded Ware enigma

The following simple dstat-figure shows the mystery of Estonian Corded Ware samples released during this spring.   There can't be any populational continuum from them to present-day Balts, including Estonians.  All thousands years older hunter-gatherer samples are overwhelmingly closer present-day Balts.  The change regarding HG ancestry can be seen in Western and Central Europe where we see a clear cut decrease of HG ancestry, obviously caused by increasing real Corded Ware and Bell Beaker ancestries.   We have to compare pure Neolithic populations against Estonian CW samples to reach parity in the Baltic area.  There is a tiny evidence about the given continuum;  Finns are closer German BA samples than Balts, giving a hint that there could be some subtle continuum.   

lauantai 24. kesäkuuta 2017

Yamnaya and Bell Beaker drift and ratio in present-day Europe

Following statistics gives an insight into how the Bronze Age Steppe ancestry transforms to a modern Northwest European genetic model and gives an idea of differences seen in Europe today.   I made free tests:

f3(Yamnaya Samara, X: Ju_hoan North)
f3(Bell Beaker Germany, X: Ju-hoan_North)
dstat(Bell Beaker Germany, Yamnaya Samara: X, Ju_hoan_North)

All results are based on around 450000 SNP's.

Results of F3_statistics were standardized to a common value 1 and also dstat-results were standardized separately to value 1.  The results show that a Yamnaya type ancestry is still significant in East Europe and the turning line from Yamnaya to Bell Beaker goes from Western Finland to Lithuania and Belarusia.   European farmer or Middle Eastern ancestry becomes dominant in South Europe leading to decreasing Bell Beaker ancestry in absolute terms.

lauantai 17. kesäkuuta 2017

Estonian Corded Ware was not Corded Ware

Despite of the common chronology the Corded Ware in Estonia was genetically a historical misstep if we believe dstat-statistics using samples of German Corded Ware and Bell Beaker cultures.  All Northern Europeans are closer German than Estonian samples.

torstai 15. kesäkuuta 2017

Shared drift with ancient Latvian, Estonian and British samples

Briefly said, shared drift of Latvian samples from Jones et al.   I have rebuild all samples using bam-files straight from the study and a new genotyping algorithm designed for ancient samples. 

perjantai 9. kesäkuuta 2017

British Viking Age samples placed on the genetic map

I got recently new samples from quite a new study, link here.   It looks more like a technical test than actual sampling for a purpose to study history, but anyway I sampled the data.  So far I have available eleven Viking Age samples from UK and have now tested them.   The data consist of around ten samples from each population, with exceptions of Swedes.  Only two Swedish samples were available for my mega-snp data base, both from the study "Genomic analyses inform on migration events during the peopling of Eurasia" (Luca Pagani 2016). The first one was from Nyköping, the second was without any place declaration.

The PCA lacks of a few Viking Age samples due to being too bad thus canceled by the outlier check.   British Viking Age samples look like to be German, but I should remind that PCA is based on dedicated components rather than genetic similarity in basis of the whole genome.   Let's see how those samples look in a formal analysis.  I have made several tests to give different views, for the reason that populations don't place in tests on one or two dimensional axis.

We see that in formal tests Swedes are closest to British Viking Age samples, followed by Irish and Scottish samples.  One straightforward conclusions could be that those Viking Age people were mixed Scandinavians and Celts/Britons. One bizarre remark:  Swedish samples are on the PCA prone to bias towards Finns and Norwegian samples show less this kind of similarity.  Still Swedish samples are closer those Viking Age samples from UK.   I have not tested this curio using formal analyses, but as far as I know this will be true in all tests.

edit 11.6.2017  12:50

German samples are from Leipzig.

torstai 25. toukokuuta 2017

PCA grouping of N1c1-haplogroup

Earlier I used TMRCA (time to the most recent common ancestor) calculation in making PCA analyses of YDNA clades, the analysis is here.  Now I use same method for grouping haplogroup N1c1.  The data was gathered from the  FamilyTreeDna's open project.   TMRCA calculation give only estimations, but  the result makes more sense because every cell in the TMRCA data is compared to every other cell.  I used 67 markers to get largest possible data.  Only a few Ftdna kits show less markers.

Download original picture here

Now I had only a few Altaic and Ugric samples.  More those samples would make possible to see the distance between Altaic/Ugric and European groups.  The result indicates three European groups:  Baltic,  Chuds and Finnish.  Actually also West Chuds are Finnish, but as far as I know it is prehistorically shared with Estonians.   The most distinct group is the Finnish one, implying local origin, despite of random distribution in North Scandinavia and Russia.

Download original picture here

The next picture shows what happen after removing Finnish clades (despite of the locations).  West and East Chuds cluster together and North Balts come close on the y-axis.  West-, East- and Central Balts cluster again.  The root group includes all samples not belonging to any named clades, but doesn't indicate any specific branch.

Download original picture here

After removing also all Chuds the picture shows more details.   We see that North Balts and Rurikids cluster together (with one classified Fennoscadinavian)  and all Balts make another cluster.  

torstai 6. huhtikuuta 2017

Estonian Comb Ceramic and Corded Ware cultures inherited to us

Thanks for the new study "Extensive farming in Estonia started through a sex-biased migration from the Steppe" I have now great new samples from Estonia dated to 4,500 to 6,300 years before present and representing local Comb Ceramic and Corded Ware cultures.   I have made dstat-analyses pointing out the comparative presence of those cultures among present-day populations. The data consisted of 11 millions SNP's to ensure reasonable coverage between ancient and present-day samples.

tiistai 14. maaliskuuta 2017

Haplotype sharing analysis, part two: Asian connections in Europe

Chromopainter is a software grouping phased data into so called chunks.  Created chunks are a practical implementation of haplotypes.  Usually Chromopainter is used with Finestructure or Globetrotter.  Finestructure reads an input coancestry matrix of individuals created by Chromopainter, which is not the best way to analyze shared chunks between populations, because it doesn't allow you to assign a coancestry connection between populations.  At least I didn't find to way to do it.  Chromopainter does it perfectly and it gives an option to use other softwares in analyzing results.  You can assign donors and recipients at population level.  This of course doesn't mean that the chunk flow goes from donor to recipient, because it is only my definition, but it defines perfectly what is common between population pairs. It neither tells us admixtures, for example the sharing between population x and Saamis tells only how much they share common chunks, not for example how much of  shared chunks are common with putative Siberians, if those Siberians even exist today.

Unfortunately my data is rather limited, some populations are well represented, some other are built only of a few samples.  In future I probably will do more similar tests and try to improve the data.  Just now I consider this step as a showcase of a new method.


edit 14.3.17 17:30

It looks like this works and it is time to play with real data.  Following small test shows how German, Icelandic and Polish haplotype references sort clearly out German and Balto-Slavic speakers, implying higher resolution than genotype data.  

keskiviikko 8. maaliskuuta 2017

Haplotype sharing analysis, part one: Europe

The following analysis was done using softwares Shapeit, Chromopainter and Finestructure.  Shapeit phasing conversion was aided by the 1000genomes V3 phasing reference.  The Finestructure report was run using chunk counts generated by Chromopainter.  Before runnig Finestructure the chunk counts file was modified to avoid "chunk leak" of population with low effective populations size.  I had earlier tested this dilemma and found that small populations being oversampled in respect to the effective population size give erroneous results due "chunk leak" towards other poipulations.  Both Shapeit and Chromopainter uses fixed effective populations size over all populations.  The remedy was to standardize intrapopulational chunk sharing to the average of all intrapopulational sharings.

Finestructure results showed also another weakness;  it is not able to treat big genetic distances in way giving readable graphic results.  For that reason I left East Uralic populations and Saamis away from this test.  I'll be back with them later.

Test conditions

- 10 randomly selected samples per population
- includes only the first chromosome
- around 40000 SNP's

Russians are from Kargopol.

Here is a link to the original gif-file, click here.


 I have also tested a new software developed by Estonian researchers called MixFit.  MixFit is a small software searching best fits using Chromopainter output.   It has a shortcoming making the fit only for three admixtures.   I tested the FinnMostCW group using same Chromopainter output as in my previous test (plus Saamis and east Uralics) and running several samples I accomplished more than three admixtures by calculating average distributions.

FinnLocal 0,3764146
West European 0,2435327
Estonian 0,2265092
Baltic 0,07271973
East FU / Saami 0,0841223

perjantai 3. maaliskuuta 2017

A short view: Were Scythians behind the Asian admixture in the European side of Russia?

As I earlier proved the Siberian admixture among Baltic Finns didn't come from East with them, it was already in Finland in the time when Baltic Finnic people reached Fennoscandinavia.  My statistics showed that rare alleles being found from Russia and Asia are in Finland just at the same level as in other European countries.

Looking closely the Asian admixture in Russia we can stretch the rare allele source to the Altay region.  How did Altaian admixture can be found in Mordvins?  Was it brought by Scythians or Mongols?  I don't know, but the fact is that it is there.

Scythian sphere according Wikipedia

For adjacent information about Mordva/Moksha see the supplementary figure 11.

perjantai 24. helmikuuta 2017

New members added to the project

Three members FI20, FI21 and FI22 are now added to the data and are now shown on following PCA plots.

Wide European PCA including Asian references

Previous PCA zoomed in

PCA including only Europeans

If you see movement in your position between the second and third PCA it is due to the difference in your Saami admixture.


I am moving on in my targets and methods and beginning to use haplotypes. and possibly rare alleles,  instead of using genotype data.  

torstai 16. helmikuuta 2017

Rare alleles show: Baltic-Finnic people are Central Europeans with Saami admixture

Speaking about Finns one of the most speculated issues have been the origin of their minor Siberian admixture.  The debate has been effusive, but in the end only boring.  Researchers have mentioned Mongols, Chinese, Nganasans, Khanties inter alia, but, as we use to say, one should not go farther than the sea to fish.  Using rare alleles, the method used by Schiffels et al. 2015 (,  we see that the Siberian admixture is credibly explained by the common history of Finnish and Saami people and the foundation of Finnish people is in this sense in Central Europe.   Of course we need to compare rare alleles of Finns and other European populations to find out who are the closest relatives for Finns and to see details.  Volga-Finnic and Eastern Uralic people show clearly different eastern admixture.  If we assume that the Finns came from Volga or Ural regions we have to explain the difference in Asian admixtures.  The simplest way to do this would be to determine the origin of the Saami-Siberian admixture and date it.  You can see this as a hint for Estonian and Finnish researchers :)

maanantai 13. helmikuuta 2017

Ancient Latvians, comparing to modern people

New ancient Latvian genomes were figured in the new study from Jones et al.   Although all new genomes show rather low quality I have now made some dstat comparisons against modern populations.  Present-day Latvians were used as a fixed point.

PCA, trying to locate three ancient samples

And three dstat figuring samples MN2, HG2 and HG3.  Most of those samples have too low quality to give reasonable results, so I have now only three results.  I tried also map original fastq files and experienced it possible, giving more available SNP's, but I decided to not use them to ensure full comparability with the study.

maanantai 9. tammikuuta 2017

Going ahead with the new data, clustering

My new data makes possible to cluster better samples according to ethnicities. It is now possible to see at least

South European
West European
East European
Finnish dwelling zone
Baltic dwelling zone

Unfortunately none of those new sample sources give reasonable South European view, which makes impossible to see inside the Mediterranean area.  With better sampling I probably could create at least Balkan, South-Italian, Iberian and Basque clusters.   It is probably now possible to classify also project individuals by PCA.

Europe, clustered by Saami, Mongolian, South-Asian and Middle-Eastern samples

Zoomed in

Europe, plotted exclusively.  You can see clearly western and eastern clusters, as well as Balts and the Baltic-Finnic group splitting into Scandinavian and East-Slavic relations.   We could see also a clearly distinct Scandinavian group with more proper samples.  Unfortunately the South European picture is fuzzy due to too few samples.  Due to the shortage of samples I narrowed each group down to four samples, except Tuscany to strengthen the southern cluster.  It is very possible that with a larger South European sampling the European west and east would diverge even more than we see now on this plot below.