DNA Connections question

Question

I tried searching and didn't find this, so hopefully I'm not re-asking a common question.

On my 6GGFCarey Clark [Clark-377] I show up under DNA Connections at ~0.78%, and there are others who show up with ~0.39%. So, I'd expect to show up on Carey's parents at ~0.39%.

But I don't. Neither does the other person who's ~0.78% on Carey.

This isn't a huge problem, but it has me wondering about the algorithm that decides which DNA testers show up on which pages, and why .39% testers show up in some places but not others.

Thanks!

Answer 1

What Profiles Get Connected to Your Test? 

• Autosomal: If you've taken a 23andMe, AncestryDNA, Family Tree DNA Family Finder, MyHeritageDNA, or Living DNA test we connect this fact to all your blood relatives out to eight degrees^[2] of separation — up to sixth great grandparents and out to third cousins.

See Help page at https://www.wikitree.com/wiki/Help:DNA_Test_Connections

Comments

What Kathie said. 

I understand standardizing on 8 degrees of separation, but since we can only test living (or recently living) individuals, that 8 degrees in terms of linear generations isn't of much use. At 8 generations back, we'd be at 6g-grandparents...and our contemporary test takers would be 7th cousins.

Any two 7th cousins will share detectable autosomal DNA about 1.1% of the time.

Eight degrees--8 meioses (birth) events up to shared 2g-grandparents and down to a 3rd cousin--makes good sense to display.

But at 6g-grandparents it really isn't 8 degrees in terms of the biology. It's 16 degrees: 16 birth events between the tested cousins. At that level, the chance of finding detectable DNA that's present in two 7th cousins and that can be accurately attributed to the same MRCA is roughly 0.6%. You're working with 1-in-167 odds.

And then you need triangulation to hypothetically substantiate that relationship. To find three 7th cousins who share some of the same DNA from the same ancestor your probability just went from 0.006 to 0.0036, or a 0.36% chance; 1-in-278 odds.

To me, it really doesn't provide any benefit to propagate DNA sharing amounts to ancestral profiles farther back than 4g-grandparents. That's a realistic threshold at which you have a possible chance of finding contemporary DNA test takers.

That WikiTree goes back to 6g-grandparents isn't a problem. It's just that we need to understand anything beyond 4g (or an ancestor's 1.56% DNA contribution) probably can't be validated by testing living people.

---

I agree and I think it’s misleading to suggest that these long-ago relationships are validated by a minuscule DNA probability

---

This is an unfortunate consequence of the way WikiTree is calculating connections. The autosomal DNA data propagates to all blood relatives within 8 degrees. This means that it goes back in time to people who couldn't have taken a DNA test and are too remotely related to possibly influence a meaningful DNA-testing match. But it also limits the cousin connections so that it does not connect our DNA to some of the living people who appear as matches in our DNA testing data.

The connection counts are an expedient way to show possible DNA connections, but they lack the sophistication that they may imply.

---

I'm with ya, Ellen. Another possible way to approach it might be unpalatable to some, but that would be to dispense with the "degrees from" calculation and, since they're computed anyway, use the theoretical sharing percentages.

Realistically, I think that could be 1.5625% vertically up the tree, which would extend to 4g-grandparents. That may be a step farther than someone could pragmatically assemble DNA evidence among living descendants, but it isn't out of the question as is 6g-grandparents.

For the lateral reach--across the tree to contemporary cousins--I'd set the bar at 0.1953%, 4th cousins...75% less than the 0.7813% of 3rd cousins.

I believe the original intent was to provide clues for potential research. Going out to 6g-grandparents doesn't really do that and gives a false impression about biological relatedness and usefulness of the information.

Conversely, limiting the lateral breadth to 3rd cousins is handy because that means the WikiTree policy of simple, two-tester "confirmed with DNA" can be applied. But finding DNA matches to 4th cousins is not unrealistic. The assumptions applied to the seminal Brenna Henn, et al., 2012 research paper arrived at an estimated average of 190 3rd cousins and 940 4th cousins: we should have almost 5 times the number of 4th cousins as we do 3rd cousins.

Limiting the vertical threshold to 4g-grandparents and the horizontal to 4th cousins is not only easier to explain and remember (4 up and 4 across), but it's actually feasible from the research-clue standpoint and has the potential to significantly increase the number of profiles carrying information in the "DNA Connections" panel.

Of note would be that the 0.1953% for 4th cousins would extend to relationships like 3C2R, 2C4R, even 1C6R. But that's okay because the biology doesn't care about when the MRCA lived, only the number of birth events separating the cousins.

---

That's a great idea, Edison. When the DNA connections were first posted on profiles, there were no percentages displayed. The addition of estimated percentages (which I believe account for nuances like half-cousin and double-cousin relationships) has created a new way to measure relationship distance.

Implementation would require having an automated system for determining whether a relationship is "vertical" or "horizontal." This might require defining various specific relationships for inclusion or exclusion. I would think that (for example) 3C2R connections would be displayed (because two living people can easily have two generations of "removal"), but 1C6R might not be (because the "vertical" dimension of 6 degrees of removal makes this a relation of hypothetical interest only).

The vertical span may, however, deserve to be increased over time as the range of birth dates of people who have been tested increased. Currently I imagine it goes from as early as about 1900 to 2024, but over time (as that second date increments) there will be more opportunities to compare DNA across "vertical" distances like six generations.