Building bridges and walls through ancestry

Copied from openstreetmap.org and modified under the Open Data Commons Open Database License.  

Bridges and Walls, Snakes and Ladders

I've noticed two perspectives within the broad scope of genealogy where it ties to population genetics.

  • Some people, those with nationalistic, right wing political views, frequently look for what divides their ancestry from others.  What defines and ties them to a historical population, or even to a land.  They may well want to prove connection to a romanticised historical group within their part of the world.
  • Others - those of a more international, liberal persuasion, instead tend to look for what unites them with other peoples alive today - what connects them within the community of humanity.

I have to confess to being more of the latter.

On Paper

I started out with a pretty well researched paper genealogical record.  A family tree.  A family history.  Researched through oral history, interviews, parish records, state records, and then on to digitalised records in more recent years.  A genealogical database of 1,570 individuals for my kids, and 207 direct ancestors recorded for myself - going back to the 1680s.  My recorded ancestry was 100% English - dominated by the County of Norfolk.  The majority of present day English perhaps have some non-English ancestry, perhaps Irish or Scottish, or something a little further afield.  I didn't find any.  All English surnames, and English denominations.  Some of those surnames however, did echo rather more ancient immigration from across the North Sea.

Autosomal DNA Testing

Autosomal DNA testing for ancestry provided a bit of a surprise.  I took a 23andMe DNA test, along with my mother, who's results I phased with to provide more accuracy.  The 23andMe Ancestry Composition analysis in standard mode didn't simply see me as English, or even as British.  It did see me pretty much as 100% European.  Not a hint of Africa nor Asia within the past several hundred years.  It saw 86% of my autosomal DNA as definitively North-West European.  However, it could only see a mere 17% as distinctly belonging to British & Irish.  So, the ancestry test of my autosomal DNA certainly agreed that I was European, NW European even, but couldn't be sure on how English or even British that I was.

23andMe Ancestry Composition in the very unreliable speculative mode rated my British/Irishness at only 37%.  The highest percentage of focus - but it saw 22% of my autosomal DNA ancestry as French / German, 1% as Scandinavian, and 2% as South European.  So considering my 100% English ancestry on paper, autosomal DNA testing couldn't really be very sure about my ancestry.  Even in speculative mode, it had 34% of my DNA as "Broadly NW European", meaning that it couldn't be sure, but somewhere in that corner of that continent.

Fair enough I suppose.  I've lost a certain amount of faith in any autosome DNA tests for ancestry to be able to pinpoint the English.  You see, even ignoring recent waves of immigration of Irish, Scottish, French, Germans, West Indians, South Asians, etc, etc.  The truth is that the English were already a very admixed population even 1,500 years before present.  Already a mixture of prehistoric populations, immigrants from across the Roman Empire, then from across the North Sea, from the Low Countries, Northern Germany, Denmark, Scandinavia, etc.  23andMe claim that their product reflects your ancestry 500 years ago.  No it does not.  It uses modern reference populations.  Genes have been circulating around the World for a long time.  Autosomal DNA tests for ancestry have really improved.  They are pretty good now for recognising a Continent - sometimes even a corner of a continent, as the source of some ancestry.  But they cannot pinpoint many populations with accuracy, and they cannot pinpoint the English.

So, my paper record said English.  My 23andMe autosome DNA test said North-West European, but couldn't even pinpoint British.  It suggested admixture.  It did however - this is important - only see me as European.  Okay, in Standard Mode, it did have a tiny 0.3% that it failed to assign to Europe, nor anywhere.  It did not see Asian.

Haplogroup DNA Testing

Haplogroups follow two narrow lines of ancestry.  The Y follows the direct paternal line, the MT follows the direct maternal.  They do not represent the bulk of your ancestry.  However, they can tell a more accurate, and longer term story.  Ancestry can be lost in Autosomal DNA within a few centuries.  In addition, it gets messed up through recombination.  Not so with the two haplogroups.  So where did mine come from?

My MT-DNA

There is an awful lot that we will know in future about our haplogroups, that we don't yet know - especially in the case of mt-DNA. However, we do know that my haplogroup, H6a1, did not originate in Europe.

H is common in Europe, and it most likely originated either there, or in South West Asia, during the Upper Palaeolithic. H6 did not originate in Europe.  It may be West or Central Asian in origin.  H6a1 has not been recovered in any ancient DNA within Western Europe.  However, it has been recovered in the DNA of the Yamnaya on the Eurasian Steppes.  For this reason, it is generally thought - based on evidence so far, to have been brought into Western Europe during the Early Bronze Age, by the expansion from the Eurasian Steppes at that time.

It isn't too fanciful - based on this evidence, to imagine that my distant grandmothers belonged to tribes of Early Bronze Age pastoralists, living on the Steppes of what is now the Ukraine.

My Y-DNA

This one has been a cracker for me.  Anyone that has followed my blog, might be getting bored with this.  I've thoroughly tested my Y-DNA.  It's not an exaggeration to suggest that it is quite likely Ancient Persian.  Based on current evidence, I believe that my Y-DNA arrived into England within the last millennia - probably between 350 and 800 years ago.  I'm still working on it's most likely route here.  I do believe that it was most likely still located in the region of Iran circa 1,000 to 2,000 years ago.  My nearest 111 STR match is to a guy in Australia who's paternal line lived in Birjand, Eastern Iran.  We shared a common ancestor around 2,000 years ago.  My terminal SNP is shared on record with only one other man so far - in the world.  He was a Balochi speaker that lives in Makran, SW Pakistan - close to the border with Iran.  The Balochi are believed to have migrated from North Iran between the 5th and 14th centuries AD.

Nomad camp, at the Zagros Mountains, Iran.  By C Whitely on Flickr under Creative Commons License.

A bit more distant, I have a Y cousin in the USA that maybe I shared a common ancestor with 3,000 years ago.  He is of Azores Portuguese descent on his Y line, but he carries a distinct STR marker that has been associated with the Parsi, who migrated to India and Pakistan, but originated in Iran.

And going further back, the Y haplogroup L most likely originated within the area of Iran and Iraq, during the Ice Age.  It would have been carried by Upper Paleolithic hunter-gatherers in that region.  13,000 years, I shared grandfathers with two Pontic Greek Y cousins, who's ancestors lived in Trazbon, Eastern Anatolia.  Maybe one Y ancestral son headed to the Black Sea, the other settled at the Caspian Sea?  The Ice Age was drawing to a close, but with a ferocity and climate instability that drove bands of people apart and into refuges at that time.

The Parsi connection keeps hinting.  They descend from Persians that worshiped the ancient religion of Zoroastrianism.  I've just seen a Y haplogroup study of men in Pakistan.  The background level of Y haplogroup L-M317 sat at 1.1%.  However, in the sample of Parsi men there - it spikes up to 13.3%.  That might not be the route however, of my Y line.  The SK1414 SNP turned up in that same study, but that was found on the Makrani Boluch man that was tested L-M317, not in the 12 Parsi men that also tested positive for L-M317.

Conclusion

I prefer bridges to walls, and that is what I got.  My paper ancestry said 100% English - much of it East Anglian.  I'm quite proud of that, but I'm equally proud of my more distant ancestors that emigrated here.  I've found North Sea admixture, from places such as the Netherlands and southern Scandinavia.  I've found a grandmother in a Bronze Age tribe of pastoralists in the Ukraine.  I've found ancient Persians, descending from hunters of Ibex in the Iran / Iraq region.  I've found distant cousins in the USA, Iran, Pakistan, Australia, and Turkey.

One species, one family.

Autosomal DNA Tests for Genealogy

First a disclaimer.  I'm very new to the whole world of genetic genealogy.  I'm not new however, to traditional genealogy, and I do have a pretty good amateur understanding of relative archaeological and anthropological discussions over the past fifty years.  The following is not meant as a critique of genetic genealogy, so much as a review, or my experience, of ancestry composition based on autosomal DNA analysis.

Let's start with my paper trail.

Traditional Genealogy

I am English by ethnicity, British by nationality, and a subject of Queen Elizabeth II (often now referred to as a UK Citizen).

My paper recorded ancestry consists of the genealogical records of:

  • Generation 1 has 1 individual. (100.00%)
  • Generation 2 has 2 individuals. (100.00%)
  • Generation 3 has 4 individuals. (100.00%)
  • Generation 4 has 8 individuals. (100.00%)
  • Generation 5 has 16 individuals. (100.00%)
  • Generation 6 has 29 individuals. (90.62%)
  • Generation 7 has 49 individuals. (76.56%)
  • Generation 8 has 35 individuals. (27.34%)
  • Generation 9 has 24 individuals. (10.16%)
  • Generation 10 has 10 individuals. (2.34%)
  • Generation 11 has 4 individuals. (0.39%)
  • Total ancestors in generations 2 to 11 is 181. (9.04%)

All 181 ancestors, reaching back to the 1690's, appear to be English born, of English ethnicity, with English surnames.  The majority of them (100% on my mother's side, and 81% on my father's side) were East Anglian, with the vast majority of that percentage being born in the county of Norfolk.  Religions recorded or indicated were CofE Anglican or non-conformist Christian.  No sign of any Catholicism, Islam, or Judaism.

Therefore it would look pretty likely, that I can claim English heritage, wouldn't you agree?

Genetic Genealogy and Ancestry Prediction

There are three aspects or avenues of inquiry, available for genetic genealogy.  First of all, the two sex haplogroups; the y-DNA, and the mt-DNA. These two "signals" are referred to as haplogroups.

  1. The y-DNA.  This follows the Y chromosome.  It is only carried by men.  It is passed along the paternal line, and only by that line, from grandfather, down to father, down to son, until the line is broken.  What a lot of people do often misunderstand, is that it does not represent 50% of your ancestry.  It does not represent all of your biological father's ancestry.  For example, his mother's father, and her brothers, although on your father's side, would most likely carry a different y-DNA haplogroup.  It only comes down an uninterrupted strictly paternal line.  Even at Generation 7 (g.g.g.g grandparents) above, it would have been carried by one out of my sixty four biological ancestors at that generation.  The other thirty one g.g.g.g grandfathers for that generation may have carried different Y haplogroups.
  2. The mt-DNA.  Although a very different type of DNA, this one works as the opposite sex haplogroup.  It is a signal that is passed down the strictly maternal line, from grandmother, to mother, to her children.  Yes, we men do inherit our mother's mt_DNA, but we can't pass it down.  Only our sisters can.
  3. The au-DNA, better known as Autosomal DNA.  Whereas the former two sex haplogroups are handy, because we can measure their mutations, and track their formation and movement across thousands of years, au-DNA really is the stuff that we are made of - all of the SNPs on our chromosomes that personalise us within the human genome.  We inherit our au-DNA from all of our recent ancestors.  Roughly 50% from our biological mother, and 50% from our biological father.  Equally, we could say on average, 25% from each grandparent, or 12.5% from each great grandparent.  However, it is messy.  At every reproduction (meiosis), it gets messed up by recombination.  Not only that, but go back much more than six generations, and it becomes more and more likely that you can lose entire lineages.  You can have no surviving trace of any DNA from for example, a particular g.g.g.g.g grandparent.

Autosomal DNA is what makes us individuals, gives us our hereditary traits.  It is passed down from many ancestors, via our parents.  However, the sex haplogroups are of interest because they can be traced across the globe, and the millennia.  As we gain more and more data - both from living populations, and ancient DNA from archaeological finds, so we will be able to track the STR and SNP mutation data more precisely.

However, what about poor old messed up autosomal DNA?  It represents our entire biological heritage over many generations. It is what we are. However, making sense of it is less easy, less precise.  Genetic genealogists are making progress, but it is far less of a precise science than either of the haplogroups.  They use calculators, that measure the segments of DNA cross the chromosomes, looking for patterns that they recognise from a number of known reference populations.  From that, these calculators predict an ancestry.  Exactly what and when that ancestry refers to, does seem to vary from one calculator to another.  There is an argument that the precision can be improved if you also test close known relatives including at least one parent.  The results can then be phased.  I'm actually waiting for the results for my mother, so that I can see my own au-DNA ancestry results phased and corrected.

So lets have a bit of fun, and see what some of the calculators suggest for my autosomal DNA, at least before any phasing with my mother's DNA.  What do they make of my 100% English paper ancestry?

23andMe.com Ancestry Composition Standard Mode

99.9% European.

Broken into:

83% NW European

17% Broadly (unassigned) European

I think that's pretty cool.  As I'm getting to know au-DNA predictions, so as I'm learning to appreciate it when they get the right continent, and the right corner of that continent.  That is more than they could do a decade or two ago.  The prediction is correct, I am a NW European.  I'm not a West African, a South Asian, or a East Siberian.

23andMe.com Ancestry Composition Speculative Mode

100% European

Broken into:

94% NW European

3% S European

3% Broadly (unassigned) European.

Whoa, where did that South European come from?  It could just be a stray incorrectly identified signal, or it could be telling me that one of my ancestors, maybe around Generation 6, were from down south!  Lets break down the prediction further.  First, the NW European:

32% British & Irish

27% French & German

7% Scandinavian

But surely I should be 100% British & Irish?  Not only 32%.  I have my own ideas about this.  I think that although 23andMe claims that Ancestry Composition only represents the ancestry of the past 300 to 500 years (the so-called migration period, as sold to USA customers), that it gets confused by earlier migrations across their reference populations, including those during the early medieval period, and perhaps even some of those during late prehistory.  I've noticed that across Ireland and Britain, the further to the east, the more diluted the 23andMe British & Irish assignment.  People of solid Irish ancestry get between 85% and 98% British & Irish.  My East Anglian results, mixed between British & Irish, French & German, and Scandinavian, are actually rather more like those received by Dutch customers of 23andMe.

As for that Southern European prediction, how does that break down?

0.5% Iberian

2.4% Broadly (unassigned) South European.

Which if taken seriously, might suggest that I have an unknown Spanish or Portuguese ancestor around Generation 6.  If I did take it seriously that is.  I wonder what my mother's test will reveal?

DNA.Land.com Ancestry Composition

This is a third party site, that you can upload your 23andMe V4 raw data to, and see what their calculators predict for your ancestry.  It has recently had it's ancestry composition revised.  What did that make of my 100% English au-DNA?

West Eurasian 100%.

I like that designation, the amateur anthropologist in me prefers that broad designation over "European".  Broken down:

77% North/Central European

19% South European

2.4% Finnish

1.3% unassigned.

What?  Why not 100% North/Central European?  Finnish?  Did some early medieval Scandinavian settlers of East Anglia bring it?  Or is it a false signal?  Misidentified au-DNA?

That darned South European kicked in again.  I'm here looking at a biological cuckoo NPE (non-parental event) at around Generation 5 or even more recent!  Did a great grandmother secretly have a South European lover?  But this South European breaks down further:

13% Balkan

6% Italian.

Oh my goodness, whereas 23andMe speculative mode suggested SW Europe - this one suggests SE Europe!  Do I have a secret Albanian great grandfather?  Or is it all nonsense?

WeGene.com

This is a cracking new third party DNA analyser.  It is based in China, and it's predictors appear to calculate mainly for a Chinese market.  It not only predicts your ancestry composition, but also your two sex haplogroups, and lots of traits and health predictions to compliment those of 23andMe.  It even tries to predict your genetic disposition to sexuality!

It will allow you to send your 23andMe V4 raw data direct to it's own calculators.  However, at the moment the website is almost entirely in Chinese (Mandarin?).  There are two options.  1) At the bottom of the webpages is a hyperlink to English, which gives, in English, a basic ancestry composition, and your haplogroups.  It does not include English versions of the health and trait results.  2) use an online translator, such as the one built into the Google Chrome browser.  It actually serves pretty well.

On sex haplogroups they give my Y-DNA as

L1.  Not bad, but they didn't make it to L1b or L-M317.

My mtDNA?

H6a1a8.  Very good.  Better than 23andMe's H6a1, and the same as the mthap program.

But this is about au-DNA, how did they do, what did they make of my 100% English ancestry?

81% French

19% English/Briton

Now, that sounds pretty awful, but on closer inspection, I'm impressed.  No South European great grandfather.  Okay, so most of my DNA has been placed on the wrong side of the Channel.  However, I know that French and English DNA is actually very close.  Recent surveys even suggest that the English have inherited a lot of common ancestry with the French during unknown migration late in prehistory.  So again - they very much got the right corner of the right Continent.  Well done WeGene.

GEDmatch.com Eurogenes K13

GEDmatch is a website that you can upload raw data not only from 23andMe, but from a range of testers, and from V3 chips as well as V4.  It hosts a number of tools and predictors - some Open Source.  Some of these predictors are for Admixture or ancestry composition.  They measure your ancestry in terms of distance from known reference populations.  The lower the number, the closer you are to their reference.  They use calculators known as oracles to predict ancestry, including mixed ancestry or admixture.

The oracles on the Eurogenes K13 and K15 calculator models have a good reputation at working with West Eurasian ancestry.  So how does K13 first, score my 100% English ancestry?

On Single Population Sharing, it rates my DNA against the closest references.  In order of closest to not so close, the top five are:

1 South_Dutch 3.89
2 Southeast_English 4.35
3 West_German 5.22
4 Southwest_English 6.24
5 Orcadian 6.97

I think that's a cracking result.  Okay, it thinks that I'm closer to South Dutch, than I am to SE English, but so close - and my East Anglian ancestry most likely does include a lot of admixture from the Low Countries from the early medieval period.  I really like Eurogenes K13.

Okay, let's now use the Oracle 4 option, to suggest admixture.  First on three populations admixing to create my DNA, what comes closest?

50% Southeast_English +25% Spanish_Valencia +25% Swedish @ 2.087456

Well that's interesting!  The SE English hit the net.  The Swedish?  Could be ancient Scandinavian admixture - but the Iberian prediction has reemerged!

On four populations admixing?

1 Southeast_English + Southeast_English + Spanish_Valencia + Swedish @ 2.087456
2 Southeast_English + Southeast_English + Spanish_Murcia + Swedish @ 2.147237
3 Norwegian + Portuguese + Southeast_English + Southeast_English @ 2.216714
4 Danish + Portuguese + Southeast_English + Southeast_English @ 2.225334
5 Portuguese + Southeast_English + Southeast_English + Swedish @ 2.230991

Oh my goodness.  K13 agrees with 23andMe AC, that I have an Iberian link.  I'm now really starting to wonder.

Let's finish off by trying K15 on my 100% English ancestry:

GEDmatch.com Eurogenes EU test V2 K15


Using Oracle for single population first, the top five closest:

1 Southwest_English 2.7
2 South_Dutch 3.98
3 Southeast_English 4.33
4 Irish 6.23
5 West_German 6.25

Okay, I'm SE English, not SW English, but pretty impressive again.

Using the oracle 4 for three population admixture, what mix comes closest to my auDNA?

50% Southwest_English +25% Spanish_Castilla_Y_Leon +25% West_Norwegian @ 1.080952

That Iberian back again!

Top five mix ups of populations closest to me?

1 Southwest_English + Southwest_English + Spanish_Castilla_Y_Leon + West_Norwegian @ 1.080952
2 Irish + North_Dutch + Southwest_English + Spanish_Galicia @ 1.111268
3 North_Dutch + Southwest_English + Spanish_Galicia + West_Scottish @ 1.282744
4 Southeast_English + Southwest_English + Spanish_Castilla_Y_Leon + West_Norwegian @ 1.295819
5 North_Dutch + North_Dutch + Southwest_English + Spanish_Castilla_Y_Leon @ 1.304939

I can't help preferring the K13 results to the EU test V2 K15 - simply because it recognises me better as SE English, rather than to their SW English reference.

Conclusions

If anyone ever bothers reading this far too lengthy post, I hope that I have imparted the following lessons:

  • Don't expect DNA Ancestry tests to pin point an actual country of ancestry.  They're not no where near that good yet.  The populations of West Eurasia, and elsewhere, are actually all mixed up, or share a lot of recent admixture.  In addition, many European nation-states are quite recent inventions.  I've seen the borders of Europe change in my short lifetime.
  • Don't expect precision.  If for example, you are an American, and a 23andMe AC test suggests only 32% British & Irish, then you could actually have 100% English ancestry over the past 300 years!  We're so mixed up, that these tests are struggling to part and identify us by nationality.
  • If you are willing to share your raw data (there are privacy issues), then have fun trying out all of these third party calculators.  It's a lot of fun as you can see.  They rarely agree.  There are other tools on GEDmatch for example, where you can compare DNA along with .gedcom genealogical files with other users - and look for shared segments on the chromosomes.  You can also compare your DNA to that of ancient populations.
  • Treat au-DNA differently to haplogroup results.  au-DNA is very interesting, and represents so much of our ancestry, if we could just sort some of the mess out.  You can partially do this by phasing your results with those of close relatives.  It is worthwhile phasing with at least one biological parent, if you can.  However, haplogroup results, provide by their mutations incredible stories over much longer periods - thousands of years.  A different kind of genealogy.  As we gather more data, and reference it also to ancient-DNA, so it will tell us more and more about two lines of descent.  Perhaps even into historical times.