41 distant cousins triangulated in MyHeritage - advice please.

Question

GS and AU (me) are '3rd - distant' cousins in MyHeritage. We share 0.2% DNA; 8.3cM on Ch-15 and 6.5cM on Ch-17. His grandmother and I share variants of the Upritchard surname and our ancestors lived about 3km apart in Country Armagh. Along with adjacent townlands in County Down, all my Irish surname variants lived in the same vicinity in 1864 (Griffiths Valuation).

We share 162 matches, but no other Upritchard's. MyHeritage offers triangulation so I took the first 88 matches and extracted a spreadsheet of how many were triangulated with GS and AU (the match list for distant cousins seems to be random and includes 5 of my known cousins from 1c to 2c-2r). 41 were triangulated matches, including 3 of my cousins. 47 were not triangulated. 1 person, my 2c-2r cousin, triangulated on ch-15. The other 40 all triangulated at the same point, close to the start of ch-17, but with varying cM (~9 to ~28 cM for 3-distant cousins of both GS and AU; (72-853 cM for AU and my 5 known cousins)). At a quick look, the 72 I have not analysed seem to have a similar % triangulated.

I conclude that GS is related to me near the edge of atDNA detection (5-6 generations; probably an MRCA born mid-late 1700's) and I'm trying seperately to find a Y-SNP from about this time.

All 41 triangulated matches must be from my paternal line. Our MRCA has a lot of descendants, a suprising number of whom have tested with MyHeritage - and probably at least an equal number have with other companies who do not offer triangulation.

Can any other information be drawn from this data ?.

MyHeritage also offer autoclusters but seemingly only down to about 20 cM, so only a few of these matches would appear in the autocluster report. (I have the report, but cross checking will be a tedious task).

Comments

You write, “ I conclude that GS is related to me near the edge of atDNA detection (5-6 generations; probably an MRCA born mid-late 1700's).” 

How are you making this conclusion? When I see a situation with the numbers you describe, my first instinct is to assume the segment goes back many generations and to move on to something else.

---

Well, I would guess that's it's the logical conclusion, based on the idea that 6 generations back is the supposed limit, and the segments are about as small as you can reasonably use.

But you can get a very small match (or even NO match) with even a 3C.

I don't know where this 6 generation "detection limit" comes from, or what it's supposed to mean. That's 4th-gt grandparents, but I share a 36.8cM segment with a half-6C that's from my 5th-gt grandfather, and I share a12cM segment with an 8C that's almost certainly from my 6th-gt grandfather (I'm trying to get some people on GEDmatch to confirm the triangulation).

---

I have usually heard that 7 generations is a limit for confirming cousins with auDNA. That fits with my experience, although it’s definitely not a hard limit.

But the OP is talking only of a segment from a triangulation group with no known common ancestors. Everyone has little sections of DNA that have come to them intact from ancestors 10 or more generations back, and this triangulated segment could well be one. The norm is that it takes an act of providence to find group of DNA-tested cousins that works to confirm the descent from such an ancestor.

---

See my answer to Frank below, that I should have ask 'Can I conclude - - '. Mathematical dilution is 100%, 50%, 25% and so on, so that when you get below 1% (7c) there are not many cM left to match. From my reading 6c seems more the accepted maximum than 7c, so to answer Franks comment, I have called this the 'detection limit' but the term now seems to prescriptive. Sorry about that. Then allow for the added complications of recombination increasing the dilution. Thus 3c can have 0 cM and 6c quite a strong cM match. I thought that weight in numbers (41 or close to 50% of all matches) would increase the probability with triangulating such small cM, but I think Frank is indicating this is not a valid conclusion.

Answer 1

It just sounds like the classic situation where you happen to have an especially long segment from especially far back, to me. I've heard them called "sticky segments", because they "stick together" more generations than is normal.

For example, I have a long segment from a 5th-great grandfather, leading to a cluster with dozens of matches on AncestryDNA. Some have appeared on GEDmatch and MyHeritage also. I think it's around 35cM long (I could look it up).

If your segment is long, than any relation that happens to have gotten any PART of it handed down will match you. Some of them won't even match each other - one has an early part of your segment, while another has a later part.

I have a number of such segments. a similarly-sized one has a similar number of matches, and I traced a number of them to a couple that married back in the 1700s in NC (I don't have known ancestors from NC). I believe that segment is probably from my 4th-gt grandmother (whose surname I do not know, but I can associate a surname with this segment, and I figure that's my best guess!

Probably one reason you have so many of MyHeritage is because it seems to be favored by Europeans.

These lengthy segments seem to be associated with mostly-male lines (I call them MML). DNA is chopped up into significantly fewer chunks when the male passes it down, vs what happens with females. So on a MML, there is apparently a much higher chance of getting an especially large segment intact.

I'd say you can tell these people on Chr17 have a common ancestor who is along your male line (or close to it), but it's like Y-DNA in that there's no way to tell HOW FAR back, from these triangulations. Even a 2C1R can share a tiny amount of DNA, so a 8cM segment can mean anything from 2C1R to 8C or even beyond!

Similarly, an really big match doesn't say much either. My brother matches a guy 78cM on MyHeritage. The segment is from our paternal line, and I was able to find an ancestor for this guy that I've run into in my research - a person who has the same name as my gt-gt-gt grandfather, shows up in the next town, out of the blue, but has to be at LEAST a 1C. So this person has to be at LEAST a 6C, which is CRAZY for a 6C. These MML "sticky segment" trash the statistics just as much as endogamy, apparently!

Comments

Thanks Frank. I should have asked 'Can I conclude - -', ie, is there weight in numbers or are 41 triangulated just by chance with such small cMs. Am I correct in reading that you suggest my statistics are probably by chance ?. In which case, GS and I really do need a yDNA match to prove our Upritchard link. Your MML concept appeals tho, as the recombination effect is greatly reduced. Say half of our triangulated matches got mostly male X on ch-23, the non-triangulated half mostly female.

---

* Quite the contrary, Alan! I'm saying that it tells you - with auDNA - about the same thing that YDNA tells you, namely that you're related, but you have very little idea how far back the connection is.

If I sound like I'm disagreeing with Barry, I don't think I am. He's emphasizing the idea, to paraphrase, that "it could be so far back that you'll never find the connection, so forget about it". While I'm saying, "sure, you're likely related", which you could call a more positive angle on it, I wouldn't be very optimistic about your chances of finding the connection, either, especially in a genealogically problematic place like Ireland. At the same time, I can't discount that this person could possibly turn out to be as close as a 3C, although the chance of that seems small.

* For one of my own cases, I have a segment from my 5th-gt grandfather up my paternal line - I've found a handful of 6Cs that share it (because I did a TON of research on them about 20 years ago). That ancestor is known to have a mother whose family was Finnish, just a few generations after the immigrant to New Sweden, in 1653.

On that segment, I also triangulate with a TON of actual, modern day Finns! These are pretty small segments. I'd say I'm definitely related to them - it's not IBC, since I don't have people from OTHER nationalities showing up. But there's NO WAY I would even ATTEMPT to find the relation. Many of these people have extensive trees, and I can't even tell how they're related to each other. They could easily be 10th cousins to me, or beyond!

Really it just makes me wonder if Finland counts as something of an endogamous population. Regardless, it gives me confidence that the segment is from the ancestor it's supposed to be from, due to his known Finnish roots.

* Getting back to your own case, I'm alittle unclear of the exact situation. I'm focusing on Chr17, because that's where you say you have a lot of triangulations. You haven't really said how big that segment is for you (vs say, a 3C on your paternal line). A paltry 6.5cM match there is pretty minimal, to say the least.But you're also saying that you know the 15Chr segment is on your paternal line also, from a 2C2R, right (again, how big is the match with the 2C2R?). At 8.3cM, that's a little better, but still pretty small.

The fact that you have TWO segments that BOTH can be traced to your Uprichards is a little encouraging, as far as her possibly not being impossibly distant, but it could also just be endogamy for that area. When I share a triangulated segment with, say, a 4C or 5C, it's not so unusual to also see a match on one or two small segments like these. Another negative is that with a 2C2R, the segment could be from one of your grandfather's other lines, off the purely paternal path.

So I'm not sure I'm really telling you all t hat much, it's just a matter of "comparing notes" about this sort of thing, where I have a paper trail that goes a bit further that what you sound like you have here.

* I don't really know what you're getting at, with that last sentence.

---

Once long ago I was a biological scientist, and my guess is I'm around 20% of the way to being a genetic genealogist, so definately NOT an expert.

For Ydna, as I understand it, FT_DNA Big-Y-700 SNP test is a quantum leap forward for tracking the male line back over millenia. atDNA stops at around 6-7 generations x 30 years per generation = 180-210 years.

For atDNA, ch-15 is an outlier - only one person triangulates. The other 40 ONLY triangulate only on ch-17, so less encouraging. I was hoping a large number of triangulated matches might dodge this issue. I tried to upload an image to show the triangulation cM range, but failed. Excluding my 2 close matches, the average triangulated segment for AU is 18.6 cM (range 8.3-50) and for GS 19.5 (range 8.3-32.8) if its any help. My 2 close matches are 262 cM (1c1r) and 72 cM (2c2r).

My paper trail stops at GG-grandfather b.abt.1875. The last sentance relates to your MML concept. The male X chromosome is not recombined, thus seemingly increasing the chances of a match. Maybe not relevant here as we are not dealing with ch-23 ?

10-August 2020 : MyHeritage have advised :

"You don't need any triangulated matches to prove you are related to GS. The fact that you have a DNA match is proof of a common ancestor. 'Low confidence' in the match overview box refers to the estimated relationship, not the match. If there was any doubt, GS would not appear in your DNA match list. The MyHeritage algorithm is set to err on the side of caution and exclude potential but uncertain DNA matches.

All people who triangulate with each other (or who appear together in an autocluster report) have a common ancestor.

Its trickier to establish how far back you are related to GS, and this is what the "Low confidence" box indicates. Since you share so little DNA, there are too many options for how you could be related. We can be fairly certain that the closest possible shared ancestor would be one of your great great grandparents, but they could be up to two additional generations back."

Presumably this description of 'Low confidence' applies to all other vendors match lists ?.

11-Sep 2020 : I have gone about as far as I can get without further help. Another 'small segment' Upritchard match has come up so the summary to date is;

GS = 164 matches, 71 triangulated on C-17; av = 15cM; range = 8-31 cM.

ML = 84 matches, 4 triangulated on C-4; av = 18 cM; range = 9-25 cM.

Maybe try to contact a few of the 128 triangulated matches and see if any recognise Upritchard 'like names' for potential Y test, but given likely 10% response I'm not very enthusiastic. Any other suggestions ?.