14 September 2020

Another Look at Ancestry's Revised Ethnicity Estimates

There have been lots of comments about the revised ethnicity estimates from AncestryDNA tests. 

Going along with the update is an Ethnicity Estimate 2020 White Paper, revised from the previous version which describes the technique as a "fast, sophisticated, and accurate method for estimating genetic ethnicity "

Like many people, my estimate increased my share of Scottish ethnicity from 35% for Ireland & Scotland previously to 19% for Scotland and 23% for Ireland, 42% in total (the uncertainty in these estimates includes 0% for both.) Many others are reporting increased Scottish ethnicity and Ancestry rushed out a blog post Why Your Latest Results Could Include More Scotland In Your Ethnicity Estimates.

That blog includes a bar chart showing that "average typical results for someone from England, they might look something like this" showing England and Northwest Europe 67%, Scotland 18%, Germanic Europe 7%, Norway 3%, Sweden 2%, Ireland 2%, Wales 1%.

My 29% for England and Northwest Europe is less than half the average typical. My Scottish is about average. My Ireland is 10 times the average, I'll have to make sure I have a supply of green for St Patrick's Day. My Norway average. 

What isn't included in the average typical result is European Jewish. Could that be Germanic Europe? If so with a Jewish grandfather I'd expect 25% which is over 3 times the average. It may be that adding the information on where you're from and comparing your percentages to the "average typical" is a useful perspective on these ethnicity results.

The ancestry blog post gives bar charts with similar percentages for people from Scotland, Northeast England and West Midlands.

Also in the 2020 White Paper Ancestry gives a table of reference panel regions and the number of samples, the top 20 with most samples are reproduced here. I wonder why Scotland has so many more samples than England and Ireland.

1Indigenous Puerto Rico4791
2Indigenous Cuba3420
3Indigenous Eastern South America2873
4Indigenous Haiti & Dominican Republic2858
5Indigenous Americas —Colombia & Venezuela2500
6Indigenous Americas —North2284
7Germanic Europe2095
10England & Northwestern Europe1326
11Indigenous Americas —Central1232
12Eastern Europe & Russia1084
13Southern Italy971
14The Balkans824
17Indigenous Americas —Mexico720

I was surprised to see the regions where most Ancestry clients would be likely to have ancestry so far down the list, and that increasing their number is not a priority — 

"We have also begun a new diversity initiative to gather DNA samples from underrepresented regions around the world in order to expand the number of regions we can report back to customers."


Donna Jones said...

John, thank you for this interesting post. The reference panel explains a lot about the changes to my genetic ethnicity. I was amazed at how small the reference panel was for Ireland and Southern Italy. It explains a lot about why I saw decreases in these two areas and a significant increase in Scotland.

Mike Stapleton said...

I don't buy into their new estimate, for me it doesn't make sense. My former estimate was 85% Irish/Scottish and it's now 70% Irish and 18% Scottish. The former results corresponding map demonstrated a Strong Ulster/Dal Riata (Scotland West Coast) affiliation which of course makes the Irish connection stronger. The new map still has the strong Ulster component but now encompasses the whole of northern Scotland and the Orkneys with a range of 0-24%. The range for the Irish 70% is 62-85%.

For myself I have a fairly good picture of my ancestors back at least 8 generations and the original estimate makes a lot of sense. Given the 6-8 generation limit of autosomal DNA I wonder if their numbers are skewed by the Irish diaspora into Scotland during and before the Industrial Revolution. This effect should also show up heavily in lowland Scotland and western England results, especially in Lancashire.

Needs more thought.

Best regards,

Mike Stapleton