Ancestry DNA test. What to expect?

In My Humble Opinion
21

This reminds me of a humorous thing I uncovered about my 5G-grandmother’s second husband. When he first appeared on the census, he was 15 years younger than her and working a day laborer. (It was after the Civil War, I guess eligible bachelors were hard to come by).

By the next census he had closed the age gap by 5 years. In his final census appearance he was not only a full 8 years her elder , he was recorded with the honorific of “Doctor” at the age of 70. The man was certainly an achiever if nothing else.

22

My question is, how big are these uncertainties? How useful in any meanigful way are these tests?

For example, I know two sisters - they have the same parents, look alike - that both took a genetic test and these are the results:

Ethnicity…,…sister 1…,…sister 2
Eng/Eur…,…58% …,.,.,…44%
Scotland …,…19 …,…,… 29
Ger Eur…,…,…13…,…14
Wales…,…,…2 …,…,.,.10
Norway…,…,.7…,…,…, 0
Eur Jewish…,.1 …,…,…,0
Sweden…,…,.0…,…,…, 3

If the results are supposed to be +/- 10%, I guess that makes this report “valid”, but I’m not sure it makes it useful. It seems like a waste of money.

I would think two nearly identical siblings would have much closer results. I suppose reporting that both of them are “100% European” doesn’t seem like you’re getting you money’s worth.

Has anyone ever done independent testing of these service?

23

No, that’s not what’s going on. Without getting into the weeds too much, there’s some probability involved. Full siblings are roughly 50% genetically related, but seldom exactly 50% related. It’s not test variation; the real genomes really can vary that much between sibs.

Each parent has an entire genome, thus between them they have two whole genomes that could potentially be passed, but each child can only get one genome. It is possible (though astronomically improbable) that each child could “draw” an entirely distinct genome from one another They would be 50% related to each parent, as expected, but 0% related to each other. Again, that’s astronomically unlikely, but possible. Usually it’s around 50%.

24

A relative of mine, 55 year old female adopted at birth, used Ancestry (and diligent followup research) to find and connect with both of her biological parents, with a “storybook ending” happy outcome.

25

I think Ancestry et. al. should ditch the “ethnicity estimate” paradigm entirely and focus on the cousin matching and improve their genealogy support, but the results you present would be totally expected even if the ethnicity test was 100% reliable. As HMS_Irruncible pointed out, siblings share roughly 50% of their DNA, and getting varying amounts of their parents “ethnicity mix” is just how DNA works.

Is the “ethnicity” part worth the money? Sure, if you see it as a form of entertainment and it entertained you.

26

These tests are just for fun, so that has to be taken into account when deciding if it’s a waste of money. Even if the results are 100% accurate what action will it let you take? I guess there might be some examples of pursuing certain genealogical information, but that is still in fun. If it’s not worth $100 of fun, then they are a waste of money.

Each child gets exactly 50% of their DNA from each parent (excluding chromosomal abnormalities), and each of the parents got 50% of their DNA from each of their parents. It’s that grandparent generation that is shuffled. So, my sister and I got 50% of our DNA from our mom, but just by chance, of the 50% that came from my mom, 65% of that was from our grandmother, but for my sister 35% of it was from our grandmother.

So, some made up numbers:
Me: 32% M-grandma, 18% M-grandpa, 28% P-grandma, 22% P-grandpa
Sister: 18% M-grandma, 32% M-grandpa, 24% P-grandma, 26% P-grandpa

As for what to believe from the numbers, this is just my gut reaction, I’d discount anything in the single digits. Each percent has to match to something, so it finds the closest match. In your example, for one sister it tosses a few percent into Norway, and the other into Sweden. What that could really mean is that they’re both related to a another group that isn’t well characterized in the data set.

You can be much more confident of larger numbers, just realize they’re not too precise. Each sister is about half English and a quarter Scottish.

As for independent testing, these ancestry sites are probably as good as it gets for this kind of thing. What’s needed is a large sample that covers lots of different nationalities, and that’s what they have. Diverse public datasets, like 1000 genomes, are much smaller, and will give worse results. Things like the UK Biobank are large, but not diverse.

27

I think it was on Radiolab where Carl Zimmer pointed out that it’s possible to share zero DNA with your 5th grandparent.

28

My younger sister just found out she’s only my half sister. Seems my mom had an affair she she and my dad were separated. She think she’s found her real father’s kids (he died 20 years ago).

29

What you’re going to find (especially as more and more people sign up for the services) is that your biological family isn’t who you think they are. Someone has a kid that “no one” knew about. Someone isn’t as genetically related to you as “everyone” thought they were (or genetically related in the way that “everyone” thought they were).

If these sites had truth in advertising, the commercials wouldn’t be about finding out you’re 1/4 Danish, they would be finding out that your Uncle Tom had a secret family in North Dakota. Surprise!

30

Lots of people find out that their biological family is exactly who they thought they were. The ones that do are overrepresented in discussion on the topic because. “No secret love children in my family” isn’t news or a mystery to be investigated, and “I don’t match any of my cousins on my dad’s side” is.

31

My family is who I thought they were, and the ethnicity piece plus the genetics piece has been very helpful in confirming and disconfirming some medical data.

32

Perhaps my friends and I are outliers then. Everyone I know IRL who has done one of the kits has had that happen eventually as more and more people submit DNA to the services.

33

You can find a lot on Ancestry that has nothing to do with DNA. They started as a research tool for documents and that still how they are most useful. The DNA part is fun but if you really want to find your family’s history you have to do a lot of research.

Through that I did find a link to Charlemagne. If you are able to find a direct link to a famous family you’ll find a lot of the work is already done. I had always heard that my father’s family had a link to Abraham Lincoln. I didn’t know what that link was. I was able find that link using Ancestry’s documents. My 7 times great grandfather was Lincoln’s 3 times great grandfather. Through that side of the family I was able to find some prominent families and find a direct line to King Edward I. Longshanks is 21 times great grandfather through his daughter Joan of Acre. Pretty funny considering the vast majority of my ancestors were dirt poor Italian, Irish or Prussian immigrants.

34

This seems like it might fit better in IMHO.

35

Yeah. Slightly different since it wasn’t the sort of thing that was supposed to be a “good” thing when I was growing up, but as an adult I found out that my brothers and sisters all “knew” that we had some non-white ancestry via our paternal grandmother - she was supposedly part “lascar,” (sailor from Bangladesh), and part something else, maybe North African. We do have dark curly hair, my Dad tanned easily, and I believed it.

Turns out we’re the whitest, most British people you can imagine. Not only are we 100% European, we’re currently 100% British and Irish.

The percentages change as 23andme - and presumably ancestry, owned by them - gets more data. They don’t separate British and Irish in the overall definition because there’s so much intermixing that it would be daft to claim otherwise.

It does break it down by areas, based on current respondents. And it’s extremely concentrated in my area of England, plus one other area I know we have contacts to, despite my closest connections being 3rd cousin - so it’s not down to all my siblings or their children taking the test.

It was sort of a shocker to realise just how little my relatives have ever moved.

36

I’ve taken DNA tests from 3 different vendors including Ancestry. The parts about where my ancestors are have surprising disagreements – majority western Europe, but for example one said mostly Scandinavian whereas the others said not. It might depend on when they consider the reference era.

All three found distant relatives that I knew of through genealogical research.

37

Italians seldom show up as having more than about 10% Italian as Italians are primarily mid eastern. I don’t think I have ever heard of anyone showing more than about 20% Italian.

38

I think it’s important to note that when these companies report you have, for instance, 20% Scandinavian, what they are saying is that their machine learning algorithms give about 20% of your genome best odds of being grouped in the same group as the genomes of their modern day Scandinavian (probably with 4 Scandinavian grandparents) reference group.

It’s a black box spitting out probabilities that are then simplified to represent a range of possible interpretations and then that is simplified again to feed you a best guess. You can, at least at Ancestry and 23andMe click through and see the higher level simplification. On Ancestry it shows you the range (e.g. 4% turns into 0%-15%) and on 23andMe you can change from the default 50% reliability to 90% and see all your specific countries shrink down and all the wide regions increase.

39

I was another one of those who found a cousin who no one knew about. She was adopted. They said she was a first cousin, but based on her age, I don’t think that’s possible (first cousin once removed maybe?). But we never were able to pin down exactly how she’s related.

My dad knew almost nothing except that he was “mostly Scottish.” Well kinda. Pretty much mixed Scottish, English and there was even some Cornish! I was happy to find I was a real “cousin Jack.” But I was more curious to see when his people came to North America and was surprised how early it was. I have no direct ancestors on my father’s side who came after 1840, and I have a number who came in the 17th C. Oh, and I could be a member of the DAR several times over. Don’t care. Found a trace of German and Irish, but generally I’m completely Western Europe, primarily UK.

My mother’s parents both came, separately, from Hungary. Interestingly, I just get generic Eastern European. Must not be very many Hungarians taking ancestry tests.

The husband of one of my second cousins was doing research for her mom (my dad’s cousin). He actually contacted some relatives and found documented history that my 3-times great grandfather’s farm outside Cincinnati was a stop on the Underground Railroad. This is the absolute best thing I discovered.

40

That’s what doesn’t seem to be made clear when they’re giving out these results. The whole notion of “ethnicity” in this context seems to me to be so cloudy as to be pretty meaningless for any practical purpose - and possibly misleading.