Sample range

Seanachai

With genetic testing are they jumping the gun when they make statements like in the article below? Wouldn't you need to have a huge sample range to give accurate information on the makeup of a nation?

For example, if for some reason everybody had to submit a sample, this could offer a much more detailed picture of the genetic origins and flip the current consensus on it's head.

http://www.dailymail.co.uk/news/article-2577003/A-million-Vikings-live-One-33-men-claim-direct-descendants-Norse-warriors.html

Ipso

There is also the no sh1t Sherlock factor; people from country that had viking invasions have viking ancestry.

Seanachai

Ipso wrote: »

There is also the no sh1t Sherlock factor; people from country that had viking invasions have viking ancestry.

Exactly, I'd say with a more comprehensive pool of samples we could even discover Mediterranean and North African links. I'm thinking about people who are more likely to send samples in and the likes of fellas up in the mountains who couldn't be bothered. I know communities where I'm from in South Leinster where I'd be very interested to see what the results are. Without being offensive they would tend to have a breeding history that is more or less confined to a certain area, at least up until recent times. If like I said most people did give a sample, then wouldn't the picture be a lot more accurate?

dubhthach

I had a draft response yesterday but got stuck in middle of something at work, so I'll start over.

First off that specific article comes from a testing company which had a "Principle" who was know for making what could be termed outlandish statements in the media. There are several such articles published particulary in the Daily Mail. The problem is they never once released a academic paper. As a result even though it's possible they had some interesting data such as in that Viking study without the rare data + numbers + multi-disciplinary team (Geneticists, historians, archaeologists etc.) producing a paper, than it's problematic. We also have the problem that this distribution map is based on modern populations and doesn't use any ancient DNA as a reference

UCL actually has whole set of pages up dealing with this company. Now I don't agree necessary with everything on that site, but I do think it's worth reading as we must remember when it comes to companies making claims that often story-telling can get in way as it's good way to sell a product. (navigation links on right)
https://www.ucl.ac.uk/mace-lab/genetic-ancestry

Anyways back to the original map, that map is specifically looking at Y-DNA Haplogroups, ScotlandsDNA had defined a number of such haplogroups of "Norse" origin, based on modern European distribution (eg. high levels in modern Scandinavia). They than use their commerical testing database to map these. When the map came out they didn't mention what specific Y-DNA haplogroups they were using, I asked them on twitter and got the following response.

@dubhthach Sorry for late reply, they're associated with norse ancestry: R1a-S200, R1a-S201, R1a-S223*, R1a-S443*, R1b-S182 and R1b-S375*

Here's another version of map that is perhaps a bit clearer.


Now there are number of unanswered questions/issues.

1. What is the total sample size
2. Do each region have equivalent sample size (relative to population denisty etc.) -- eg. not a case of 10 men in Connacht vs. 100 men in Orkney etc.
3. What is justfication that set list of Haplogroups are potentially Norse?
4. Could some of these haplogroups be found in other continental Germanic groups (eg. could some of them be linked to Anglo-Saxon's and not Vikings)
5. Was there any input from Historians etc.

Now it should be pointed out that as far as I know we have no ancient DNA from either the Viking period or the proceeding Scandinavian Iron ages/"Nordic Bronze age" when it comes to Y-Chromosomes anyways. (I could be wrong on this).

Now there are some probable nuggests of truth in this map, so for example both Orkney and Shetland were part of the Kingdom of Norway until they were pledged as security on Dowry in 1468. A distinct language called "Norn" which was closely related to Faroese (and Icelandic) survived right into the 18th century with last speaker probably dying in the 19th century.

Seperatly the Outer Hebrides which is now a Scottish Gáidhlig stronghold was part of the Norse Kingdom of the Isles, it would appear that it actually underwent gaelicisation probably due to influce of the "Gall-Ghaeil" (Norse-Gaels), so potential language shift of Pictish -> Norse -> Middle Irish. This is area of world that would late produce the Gallowglasses who are clearly a hybrid culture.

Likewise it has been theorised that Manx Gaelg is result of Old/Middle Irish been introduced to the Isle of Man by the Gall-Ghaeil, it likewise was part of the "Kingdom of the Isles" and technically for while under the control of Kingdom of Norway.

It is notable that we thus see elevated levels in Orkney/Shetland and Outer Hebrides and Isle of Man. Particulary as we know that Anglo-Saxon influence was next to nothing in most of these areas before the Norman conquest of England.

The issue that arises of course is without exact numbers the map is meaningless, as Seanachaí implies we don't know if they have only 10 or 100 samples from Connacht, if it was 10 samples than the result could shift radically if another 100 men were added into the analysis.

What I would say though is that most large scale scientific studies have pointed to the likes of R1a been rare in Ireland, the specific strains mention above are often seen in men with surnames linked to the Ulster Plantation, or potential Gallowglass origin.

To give you another example of map from this company here is their map for M222.



The basic pattern appears to match previous academic studies such as Busby (Royal Society) and the TCD study (Bradley etc.), however there are at least two issues I can see:

1. How big is the dataset -- how many men for each province in total and what number are M222+
2. The boundaries are based on modern provinces and bit too coarse grained. Ideally Dublin would be spilt out and that Leinster would be spilt in two to try recreate the historic province of Meath. Also I would spilt out Greater Belfast region.

M222 has been hypothesised to be linked to Niall of the Nine Hostages, Trinity published a paper back in 2006 suggesting this which was picked up with media, as a result some testing companies will stick a badge on someone profile when they are M222+ (or fit the STR pattern) saying "Niall of Nine Hostages" (with Irish flag). Of course problem with this is that (a) we don't know if Niall was real (b) if he was we don't have a body to prove he was M222 (c) even in pseudo-historical narrative he had several brothers who also produced dynastical groupings.

Busby study of Ireland had the following dataset

n= 476
M222 = 105 eg. 22.06%

That was made up of 9 distinct sample sets, some big some small. Here is breakdown of them where I've combined some to give something close to province setup.

Connacht -- Galway and Ballinrobe
n= 83
M222 = 25 (30.12%)

Ulster -- Magherfelt, Belfast, Virginia (Cavan) -- heavily biased to Belfast
n=109
M222 = 35.77%

Leinster -- Dublin and Carlow (heavily biased to Dublin)
n=149
M222 = 29 (17.34%)

Munster -- Thurles and Newmarket (Cork) -- heavily biased to Thurles
n= 111
M222 = 11 (9.9%)

What should be noted that we see similar pattern as the map, though it's arguable that we could probably take the Virginia Cavan sample out of Ulster and use it to give us an idea of historic area of Meath, after all it would have been part of Meath before been conquered by Bréifne.

What's evident is there does appear to be distinct cline when it comes to this Y-Chromosome haplogroup. Which potentially has an echo in Irish history in the period 400-1169.

dubhthach

It should be noted that the Y-Chromosome only makes up about 2% of a man's entire genome. One reason why it's attractive for genealogy is that due to fact it contains genes that encode for maleness the bulk of it can't recombine with the X Chromosome, as a result it's basically passed down unchanged (unless a mutation occurs) between father and son, in many ways similiar to how surnames are.

However it's quite possible to have a Western European Y-chromosome but on level of rest of your genome to be quite distinct. A good example of this is in African Americans were anywhere between 20-30% of African American men carry European Y-Chromosome lineages -- a tragic outcome of the slavery.

riffmongous

What is the effect of the margin of error work in these kinds of studies? It's not my area but just from a quick look you have fairly small sample sizes and a lot of small results, how meaningful are they really?

Seanachai

dubhthach wrote: »

It should be noted that the Y-Chromosome only makes up about 2% of a man's entire genome. One reason why it's attractive for genealogy is that due to fact it contains genes that encode for maleness the bulk of it can't recombine with the X Chromosome, as a result it's basically passed down unchanged (unless a mutation occurs) between father and son, in many ways similiar to how surnames are.

However it's quite possible to have a Western European Y-chromosome but on level of rest of your genome to be quite distinct. A good example of this is in African Americans were anywhere between 20-30% of African American men carry European Y-Chromosome lineages -- a tragic outcome of the slavery.

From what I've read you also have to factor in the non-paternal factor also, a line of men with a particular surname may not actually be from that group if the original ancestor wasn't who it was presumed to be. If Mrs Doyle becomes pregnant but it's actually by Tommy Kinsella the traveling bard, as opposed to Mr Doyle, but the boy is named Doyle and no one is the wiser, then his line isn't actually Doyle.

dubhthach

riffmongous wrote: »

What is the effect of the margin of error work in these kinds of studies? It's not my area but just from a quick look you have fairly small sample sizes and a lot of small results, how meaningful are they really?

I never did pay attention to Statistics as a subject (my bad), it's good question one which I can't supply. Generally academic papers will section where they would mathematically model a population etc. However it would indeed be nice if there was a study with 20k Irish men in it (with at least paper trail back to Great-Grandfather or earlier in Ireland)

I recall on another forum someone did an analysis of some of figures from the Ireland yDNA Project (7k members), what they did was look at men who had done some sorta SNP testing (eg. they had tested for M222 etc. as oppose to just getting prediction based on STR results).

The figures were along lines of following:

This resulted in 1675 samples.

Highlights include:
R1b = 81%
L21 = 60% (eg. subset of wider R1b)
I =13.5%
U106 = 4.8%
DF27 = 2.75%
CTS4528 = 2.1%
U152 = 2%
J = 1.6%
R1a = 1.5%
G = 1.1%

The caveat above is that there was a cohort of close on 200 men who had only done partial SNP testing, so hard to say if they were L21+ etc. if around half of them were it would bring total up to 65% not million miles off the figures from Busby.

The M222 cohort was 343 men out 1675 which works out about 20.47% (Busby had 22% on sample set of 476)

In case of the Busby (Royal Society, 2012) paper stats above I'm just pulling them out of Supplementary information which is made available. Of course larger a dataset the more accurate any result will come out.

The main reason though I spilt them by province was to contrast with the ScotlandsDNA map which we have no academic paper or raw stats on! The original article that Seanachaí posted was a PR exercise by same company.

Also I wanted to compare it with the Trinity College study from 2006 which produced this map:

A Y-Chromosome Signature of Hegemony in Gaelic Ireland
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1380239/



Now that study was based on surname analysis as well, seeing if Uí Néill type surnames were enriched for specific marker set.

From a historian point of view there are two interesting essays looking at this which were published a couple years ago. (in Seán Duffy (ed.), Princes, prelates and poets in medieval Ireland. Essays in honour of Katharine Simms (Dublin, 2013).)

Medieval Irish genealogies and genetics,
by Bart Jaski
Medieval Irish genealogies and genetics, in Seán Duffy (ed.), Princes, prelates and poets in medieval Ireland. Essays in honour of Katharine Simms (Dublin, 2013) 3-17.
https://www.academia.edu/2563825/Medieval_Irish_genealogies_and_genetics_in_Se%C3%A1n_Duffy_ed._Princes_prelates_and_poets_in_medieval_Ireland._Essays_in_honour_of_Katharine_Simms_Dublin_2013_3-17


Interlaced scholarship: genealogies and genetics in twenty-first-century Ireland
by Catherine Swift
https://www.academia.edu/3363365/Interlaced_scholarship_genealogies_and_genetics_in_twenty-first-century_Ireland

In Genetic Genealogy community obviously there is strong interest in the interconnections between surnames, so for example in traditional Irish genealogies you will see that "Surname X" is said to be related to "Surname Y". Also you will often see that there are multiple independent occurences of a surname. To give a simple example a good surname to look at is McManus

In the medieval genealogies there are specifically two McManus families mention, in both cases they arise as distinct surnames in the 14th/15th century and branch from major ruling families.

McManus of Connacht -- branch of O'Connor's, descended from Maghnus Ua Conchobair (son of Tairrdelbach Ua Conchobair -- High King of Ireland)

McManus of Fermanagh -- branch of the Maguires, descended from Maghnus Mag Uidhre (great-grandson of Donn Óc Mag Uidhre d.1302)

So genetically you would expect to see a minimum of two distinct genetic clusters within the surname reflecting (a) two different founding lines (b) that both parents branches were historically important. Along with that you would expect to see men who didn't match either due to either adaption, surname change (assimilation of rarer irish surnames to larger more common one etc.) or NPE's (Non-Paternity Event -- eg. what Seanchaí is talking about with Doyle's and Kinsella example)

So we see one branch which is M222+ (more sepcifically A259+) which has members with known ancestry in Roscommon, these men show STR matches with men bearing surname O'Connor. They also happen to match with a relative of the current O'Conor Don (verified as M222+).

Another branch is seen where the men fall into a genetic cluster with men bearing surname Maguire and other related surnames (McAuley for example),

Thirdly we see men who don't match either group and who either show up as singletons (Eg. without other McManus matches) or in smaller clusters (denoting a shared common ancestry).

dubhthach

Seanachai wrote: »

From what I've read you also have to factor in the non-paternal factor also, a line of men with a particular surname may not actually be from that group if the original ancestor wasn't who it was presumed to be. If Mrs Doyle becomes pregnant but it's actually by Tommy Kinsella the traveling bard, as opposed to Mr Doyle, but the boy is named Doyle and no one is the wiser, then his line isn't actually Doyle.

I hope Tommy didn't impress Mrs. Doyle with the feats of the Táin

Yes Non-paternity is often issue when looking at results. In case of Autosomal testing (such as Ancestry/FamilyFinder) often it gets noticed when someone ends up with cousin matches which don't match the papertrail.

In case of Y-DNA (our Kinsella/Doyle scenario above), if a sufficiently thorough Y-DNA test is carried out than what will often happen is following:

1. Young Doyle at age of 18 tests to 111 STR's ($300+) and orders a SNP pack (M343/M269 for argument sake 00 ~$119 to nail down his haplogroup)
2. Instead of showing lots of other Doyle's as matches he see men with surnames such as Kinsella, McMorrow, Kavanagh, Murphy (STR matches)
3. SNP testing shows he falls into a R1b-Z255 which is heavily concentrated in surnames of Gaelic Leinster origin

This raises question of (a) is he really a Doyle or (b) have no other Doyle's tested and thus no comparison.

Now according to this link there are 268 Doyle's in FTDNA database:
https://www.familytreedna.com/surname-search-results.aspx?sType=eq&Searchname2=Doyle

As oppose to 25 Kinsella's (though you could add in closely related surnames of Cavanagh, Murphy etc.)

Needless to say genetic genealogy should always be seen as an adjunct to existing tools of genealogy, so for example if you have a papertrail back to particular individual and you believe that another individual is a 4th/5th cousin (also descended from this particular ancestor) a Y-DNA test (both of ye been male direct line descendants) -- should show a match, if it shows a distinct non-match (eg. not possible you share direct male ancestor in last 1500 years) than it would point to either a "break in lineage" (eg. NPE) or that two individuals are from different lines and don't have same titular ancestor.

What I would say is you can also do testing and show next to no matches, particularly if you come from a surname that wasn't one of the "Great and Good" in olden times

dubhthach

One thing I forgot to mention is that when it comes to population analysis on a European wide scale you often see mention of a cohort called "Irish Trinity Student Controls" -- this cohort was recruited out of the student body of Trinity college based on self reported ancestry/ethnicity. It's often used in academic studies on disease risk etc.

http://www.pnas.org/content/110/2/588.full.pdf?with-ds=yes

Trinity Student Study. A cohort of 2,524 healthy, ethnically Irish individuals,
attending the University of Dublin, Trinity College, with ages between 18 and
28 y, was recruited over one academic year in 2003–2004 (21, 23). Ethical
approval was obtained from the Dublin Federated Hospitals Research Ethics
Committee, which is affiliated with the Trinity College, and reviewed by the
Office of Human Subjects Research at the National Institutes of Health.
Written informed consent was obtained from participants before recruitment.
Details of the sample collection, genotyping, and data-cleaning process are
described in SI Appendix.

As the majority of the TSS subjects were unrelated, population structure was
investigated using a single-step PCA involving all 2,310 subjects and 210
HapMap reference samples and a pruned set of ~117K SNPs, obtained from the
initial 758K SNPs using the same pruning criteria described above. All 2,310
subjects were verified as having European ancestry, consistent with their
reported Irish ancestry.

I believe this cohort was used in the recent paper on Irish Travellers as a reference.