Possibility of triangulating using <5cm segments?

Question

Dear Wikitreers,

I've been puzzling over an atDNA conundrum and am hoping someone else will have had a similar experience and be able to provide a case study one way or another.  It's generally agreed that atDNA matches in the 2-4 cM range are overwhelmingly IBS rather than IBD, with false positives accounting for something in the range of 67-85% of all segments (see the relevant isogg page).

Presumably, though, this means that 15-33% of these small segments are IBD.  I've been looking at a couple of instances recently where three or more individuals share a segment of 2-5 cMs and also have paper trails linking them back to a single ancestor in the 1650-1750 range.  My conundrum, though, is: are these shared segments really IBD or am I looking at something else on a population level or otherwise?  I haven't had much luck finding any articles which address this sort of corner-case scenario, but would be very glad of your insights.

All best wishes,

Kelsey

Answer 1

Blaine Bettinger did a little survey on Facebook where he asked who share segments with him in a 3 to 5 cM range. There were well over 100 responses and 87% said yes.

You just can’t use segments that small for genetic genealogy.

Comments

If you're talking about the survey he ran last week, I participated in it.  I shared 16 3-4 cM segments with him.  By the end of the survey he had over 1000 participants, and 94% matched him on at least one small segment.  When he did spot checks of some of the people in the remaining 6% who said they didn't match and walked them through the procedure, it turned out they did actually match him.  So I'm very very skeptical of small segments.

---

I completely agree.  The basic fact that an individual shares a small segment with another individual means effectively nothing.  What I'm more curious about is when several individuals share the same small segment and also have a common ancestor on paper.  It's not clear to me whether, in that limited scenario, the segments are still a matter of chance or whether they're a genetic indicator of the paper ancestry.

All the best,

Kelsey

---

It seems like they could be, but I don't know how you would know for sure.  Maybe if you were able to demonstrate a "chain" going back multiple generations sharing the same ever-shrinking segment?  So if, for example, you have documented 3rd, 4th, 5th, 6th, etc. cousins on that same line that share a larger segment that includes this small fragment, and if your distant matches have the same situation for their closer cousins.  That to me would be more suggestive of it being IBD.  (Caveat: I haven't totally thought this idea through and there's probably something wrong with it...)

---

It is not impossible that the segments are from that common ancestor rather than just an interesting deal from DNA deck. How can you tell, though?

If all 3 of you have a good paper trail and you really want to prove it through DNA then you could always get more creative.

I have a 5x great grandfather with a ridiculous number of living decedents, most of whom seem to love to DNA test. I have about 30 distant cousins on that line and good DNA networks among them but the segments are all 7 to 12 cM and there is likely some endogamy in that colonial American group.

So I noticed that a few of those distant DNA cousins with the 5x great-grandfather's surname had YDNA tested. I found a 3rd cousin with that surname who I strongly matched with aDNA and had him YDNA test. Sure enough, he was a match with those 6th cousins.

---

Davis, great minds think alike!  Our posts overlapped.  Same general concept.  

---

lol yup!

---

Thanks both for your thoughts!  The idea of a chain seems like exactly the right way to go about trying to answer a question like this.  Now to find those intermediate cousins . . .

All the best,

Kelsey

---

I'd like to point out that Blaine's "3cM" survey measured something else entirely. He ran his survey looking for people who matched *any* 3cM segment of his...  While Kelsey is asking something very different:  How reliable is a *triangulation* where 3 or more people share the *same* 3-4cM segments... *and* throw in the paper trail.  Yes, there is the question of Endogamy, but while the chance of sharing a random 3cM with another person is *high*, the chance of sharing a *specific* 3cM has got to be significantly lower.  Low enough to adjudicate a positive match?  I don't know.

---

Thanks, Brian.  You've posed exactly the question I was trying to ask!

Answer 2

Blaine Bettinger did a little survey on Facebook where he asked who share segments with him in a 3 to 5 cM range. There were well over 100 responses and 87% said yes.

You just can’t use segments that small for genetic genealogy.

Answer 3

First of all the answer is yes!

Some people here don't seem to understand what triangulation means. If you find several people for whom you have a paper trail and they do share an overlapping segment of DNA on the same chromosome than it means that they have all inherited from the same common ancestor.

As you rightfully stated not 100% of all small segments are IBS or IBC, some are still IBD. So I have to fully disagree with Davis Simpson's answer below.

It's actually pure logic because DNA doesn't create itself. It's copied over and over again and the number of recombination events per generation is pretty clear. Hence DNA segments, no matter how small they are can be IBD.

Just because the majority of not well informed people says doesn't make it a fact or right (in the same way that not always the majority of family trees at Ancestry.com are the correct ones, just as a side note).

Actually the smaller the inherited segment of ancestral DNA is, the higher is the chance that it will survive for many generations intact. Again, proven by scientific research because the number of recombinations across the whole genome is relatively stable.Think of it like the chance to be hit by lightning. If your as wide as a football field your chances are much higher that lightning will hit you somewhere vs an ant that is somewhere on the football field.

Comments

Dear Andreas,

Thanks for this and I see what you mean.  Certainly it makes intuitive sense that  this should be the case (while I also recognise that intuitive sense doesn't always equate to truth!).  As, perhaps, a more concrete example from my own ancestry, I've noted that four descendants of Kjell Christoffer Barnekow (1663-1700) and Margareta von Ascheberg (1671-1753) all share a 2-4 cM segment on chromosome 8 between 41/42m and 52/53m.  As Brian said a moment ago, if they simply shared *some* or *any* small segments, that would be meaningless, but I find myself very tempted to suppose that a scenario in which multiple individuals descended from the same CA also share a small segment would indicate that that segment is IBD.  It would be interesting to test this further and see if that holds or whether we could still be looking at a case of random, fortuitous recombination.

All the best,

Kelsey

---

I agree that 3 cM segments can survive for much longer intact, absolutely. My mother has an 8 cM stack that I'm pretty sure has been intact for many hundreds of years through endogamy.

I don't think all 3 cM matches are because of pure random chance. Many, many such segments, and even much larger ones, are present in whole populations through endogamy.

So if they did inherit that tiny segment from a common ancestor, who says it's from that one in their paper trail? If they have ancestors from the same region, they likely have more ancestors further back through other people than through that common paper-trail ancestor and that 3 cM segment is pretty widespread in that population.

I really can't see how you can use a 3 cM segment as proof as anything.

---

Dear Davis,

I should clarify up front that I'm not trying to use a small segment as proof of a relationship - rather I'm wondering whether a paper trail in the scenario I've described might point towards the likelihood of shared small segments being IBD.  I agree as well that at that number of removes the possibility of the individuals in connection descending from multiple CAs is - depending on circumstance - quite large.  That said, in the Barnekow example I gave earlier, one of the descendants who shares the small segment at chromosome 8 is of entirely Dutch ancestry back to the beginning of the eighteenth century save for a single Swede (grandchild of Barnekow).  That would seem to eliminate much probability of that match sharing much common ancestor with the other matches, all of whom have mostly or entirely Swedish/Baltic-German/Russian ancestry.  Still, I agree it's all very conjectural and at the edge of atDNA's interpretability!

All the best,

Kelsey

---

Hi Kelsey,

If you are just curious about the possibility, you might want to use a tool like DNA painter to trace that segment further. See if you can trace that segment back through the generations.

My father shares a 7 cM segment with a cousin on his paternal line with whom he shares a common ancestor in Cornwall. However, I don't share that segment so I must have gotten it from my Dad's other chromosome.

I then figured out that I had matches on my Dad's paternal line on segments that included that location, which means my Dad's match with the Cornish cousin was actually on his maternal line. Turns out they were related somewhere in ireland.

Even if you can't use the 3cM segment to prove a relationship, I agree it would be interesting to see if it came from a certain population.

---

The question was if you can triangulate segments under 5 centiMorgan and my resounding answer was yes.

As to who is the common ancestor, how many years back he/she is and so on is a totally different question.

But again, if such a segment triangulates (and the rules are the same for a 12 cM or a 4.5 cM segment) then it's IBD and that means it's inherited from a common ancestor. Now which way it has come down (I agree there can be several paths) can be explored by coming the triangulation results and the family tree information. It is very likely, that this small segment isn't the only segment that some of the descendants of said common ancestor are sharing with some (most likely not all but usually over 70=80% of all people in the TG) of their DNA cousins in the TG.

Just to repeat it again, even if there is only a 0.1% chance that a segment is IBD it doesn't mean that all segments are IBS/IBC!

---

Kelsey, you write “I should clarify up front that I'm not trying to use a small segment as proof of a relationship,” but the subject of your question is about using small segments for triangulation. The short answer is that you can’t. This comes up again and again; one of the largest DNA threads on G2G is about shared tiny segments among people with Pennsylvania German ancestry, and it’s almost all nonsense.

Answer 4

If you've got a robust paper trail then that is much more reliable than DNA evidence based on low cM matches, irrespective of any triangulation of them.

I have two surname projects that I'm working on at the moment and have been investigating whether I can use small matches to augment the skimpy paper trails. Both studies have about 50 donors and a high expectancy of relationship based on an uncommon surname coupled with high geographical overlap.

I originally used 4cM as the low end cut-off, and within each group (comparing ~50 people with each other) there are a very large number of matches at that level, but with no pattern in them that I could pick out.

Within the groups there are some parent-child couples so I was able to do simple phasing of those against everyone else.

At the 4cM level 93% of the matches disappear on phasing.

At 5cM, 88% disappear.

At 6cM, 60% go.

This is in line with the data published by Tim Janzen (on the ISOGG page you mentioned) which goes one step lower than I have and says that in his dataset the false hit rate at 3cM is 99%.

So at the 2-4 cM level the false hit rate is more like 93-99.9% than 67-85%.

That said, within my two studies I do have two low level triangulations that I think are true. One is at about 4.5 cM for 4 donors, bolstered by one of the pairs matching at 8cM which is what makes it more reliable. The other is at 5.0 cM and I think that is probably reliable as two of the donors are phased kits.