Saturday, July 9, 2011

Let's consider the regions

In the last post I tried to make a start on ordering our data and representing it graphically. This helps define some jobs we have.

First, we need definitions of regions which are big enough to be statistically significant. Second, we need to define haplogroups in the smallest possible clades that are still clear and meaningful. In this post I will just discuss the first challenge.

The obvious aim of the project is to eventually have a big data set for every traditional county in Britain and Ireland. But at this time, at least just using our own data, for some counties we have very little.

Therefore, I have adapted some fairly standard ways of defining regions, joining together counties until our data is rich enough. Unfortunately, for now this mean Wales is one region, and Ireland and Scotland are defined reasonably broadly. It will be interesting, of course, to compare our results in these regions to those of Wales DNA Project, the Ireland Y DNA project, and the Scottish DNA project. Eventually maybe we can work with them and other projects to develop bigger and better descriptions of the genetic diversity of Britain and Ireland.

I've used the so called "NUTS 3" regions of the republic of Ireland,

...but the only one I did not merge with several others was the Border Region, which surrounds modern Northern Ireland, which is of course part of the United Kingdom. Also, I re-united all of Tipperary, which today is in North and South parts. Both parts are in my "Western Ireland".

For Scotland, the best data we have is for the counties south of the narrow part between the Firth of Forth (near Edinburgh) and the mouth of the River Clyde (near Glasgow), the area which includes the bulk of the modern population. I was able to split this into two near and traditionally meaningful parts. On the east is the area between Edinburgh and England which was settled early by Northumbrian Anglo Saxons. In modern parlance, this is called the region of Lothian and the borders. On the west, from Glasgow down to England, was a Welsh speaking kingdom. In modern terms this region is called Strathclyde (the northern part) and "Dumfries and Galloway".

For some modern Scottish region names see Wikipedia.

For the northern part of Scotland I could only get big data sets by uniting most of the highlands into one region. Perthshire, and the "lowland" shires to its south and east, I have been able to separate out.

For England I've used modern definitions of regions,
...except that I have taken the opportunity, given our data, of inventing one new region which I call "East and North of London". This includes the Thames Valley counties of the South East England Region, and the inland part of the East of England region. This neatly allows us to split out the coastal region of East Anglia, which is of course interesting for anyone interested in trying to find signs of ancient movements of people.

There is also a good East Anglian DNA project we can compare to. I invite help and comments concerning how this project's data compares to those of other projects.

For now I have not yet attempted to develop anything with the various remote islands. We have some data but not much yet. (The Isle of Wight is however part of the region south of London. It is very close to the mainland.)

So here is a map:-

Sunday, July 3, 2011

A first quick effort

Let's see if this works! Please click on the picture below, and start those comments rolling in. What should we do to improve this graph, and what does it mean?

Note: the haplogroup assignments are by FT DNA and include their predictions.

The data here is just our project data, but:-

  1. I combined STR and SNP data (two files that FT DNA's controls create for admins) into one sheet, lining up the data.
  2. I organized the countries and counties and invented the regions, which means I also sometimes corrected what people had down as their COUNTRY of origin, because they are often wrong. A lot of people apparently don't know which country some counties are in, or else they were hedging bets. I have assumed their COUNTY information is correct, because that is the data we always push people to double check in our project.
  3. I ran my own haplogroup prediction using Whit Athey's tool, but I have not used that information much yet.
  4. I removed everyone without a pedigree to a county. Maybe that was the most important step!
  5. I created a frequency table using pivot table functions, which I have e-mailed already to both of you, and a graphic representation of that frequency table, which now appears on the blog. (Two work sheets in this spreadsheet.)
  6. I created a short version of the haplogroup names so that they all line up and look the same, not depending on the SNPs tested.
And here is the data:-

And here are the regions I have used, in order to get big enough data sets, of people with pedigrees back to old counties:-

Here are a few first remarks:-
  • G levels highest in Wales in the northern part of the Republic of Ireland. Remember that people are now saying this is a Neolithic farmer (pre Celtic) marker, based on the relatively large number of G men found in old archaeological sites.
  • I2a in interesting patches: western Ireland, most of Scotland, NE England, and the extremity of SW England, but apparently almost absent in many areas neighbouring on these, like SE Scotland, NW England, and the counties neighbouring the extremity of the SW of England.
  • I2b almost invisible in southern Ireland and Wales, but high frequencies in southern Scotland, northern Ireland and also common in most of England.
  • I1a pretty common everywhere except in western Ireland, but if it is Anglo Saxon you would expect it to be higher in SE Scotland?
But I have to say that I haplogroup prediction from FT DNA and also from Whit's predictor can probably be improved upon. I have contacted the obvious people: Jim Cullen and Ken Nordtvedt. I haplogroups perhaps deserve their own post in the near future.

What members should do


Make sure that on your FT DNA personal page you have good up-dated information about the country of origin and county of origin of your male line ancestors.

I can not emphasize enough that missing information or inaccurate information in these areas makes it hard to achieve the core aims of the project.

We should actually do something

My thanks to Ken Nordtvedt for writing to ask for data from this project in a specific format he wanted. It is a useful push for action.

It has to be said that we are still in a position of having a big mess of data which is difficult to use. Ideas on how to improve this are welcome.

One big issue is that we have an enormous number of participants who have no known male line connection to the British Isles at all. Many even know they are from somewhere else. The sheer number of such members does make all jobs difficult in my opinion, although I understand that people want to have their Y DNA in the database just to feel a link to the project.

Keep in mind that there is also an even bigger amount of people who are members but only believe that their ancestors were from the British Isles, not which country, or perhaps they know which country, but no more. And I am sure all experienced genealogists agree with me that we can expect most of these people are reporting what are essentially guesses. (Many family stories are just the guesses of a previous generation.)

Anyway, enough complaining. One thing we have long aimed to do is to create haplogroup frequency data in a more user friendly format. I am going to get to work and at least do some preliminary work.

To start with I've just made an excel sheet where I've deleted all people who have not reported a clear county of origin in Britain and Ireland. That makes it much easier! Only about 1500 people!

So I would like to ask for opinions on how to divide up the populations in terms of haplogroups? Many participants have of course not been tested for any SNPs, while some have been tested for all the latest new ones.

I am supposing I'll need to run part of the data through a prediction program like Whit Athey's. Should I also ignore all people with less than a certain number of markers?

Of the approx 1500 people who know a county of origin in their male line, a bit more than half only have a predicted haplogroup the way they appear in the FT DNA data. About 740 have had real SNP tests.