How does one triangulate matching DNA to confirm an ancestor?

Question

Woohoo! I've found an 18th century ancestor who is alleged to be an American Revolutionary War veteran AND there are three Wikitree members identified as possibile DNA sources.  Great. Now what do I do????

All three DNA kits were processed through FamilyTreeDNA. The first two are myself and my mother. Does that mean we have to get at least one other person not so closely related? I haven't yet met the third person. Hopefully, she will turn out to be some sort of distant cousin.

What are the mechanics of comparing the DNA results? I've never done this before. Exactly what information are we to identify on all three test results?  BTW, what does the report's inclusion of "mutations" tell me?  I feel like a school girl who hasn't prepared for an exam.

Thanks in advance.

Comments

Carrie, as you got the ball rolling ....

You can't triangulate through FTDNA's website without the help of one of the other two matches (they need to tell you where exactly they match the third person). We had that (heated) discussion already on the difference between ICW and triangulation.

Unfortunately FTDNA is still not providing us with the right tool to do triangulation without help from other people, only 23andme does that.

---

Thanks for that clarification regarding the FTDNA utility, Andreas.

---

Not quite true.  First, you can use the Chromosome Browser to determine if the other two people match you on the same chromosome in the same location, with a reasonably long overlapping segment.  If they do, then you can use the FamilyFinder Matrix to see if they match each other.  If they do, then they are almost certainly triangulated.  You can confirm the triangulation by checking with at least one of the other persons who matches you to make sure they also match the third person.  (This is the equivalent of going from a derivative record source to an original record source.)

---

Gayel,

Which is exactly what I wrote, you need help from one of the two matches to tell you where they match each other. There is no way (due to privacy concerns at FTDNA) that any of the tools that FTDNA provides you can tell you where B and C are matching each other (you can only see THAT they match each other).

So your "Not quite true" is wrong! This statement I wrote is true: "You can't triangulate through FTDNA's website without the help of one of the other two matches"

You wrote: "If they do, then they are almost certainly triangulated" - well, certainly isn't 100% and that's what triangulation when done properly can achieve.

You also wrote: "You can confirm the triangulation by checking with at least one of the other persons who matches you to make sure they also match the third person." - no, this isn't confirming the triangulation as you're not checking where they match each other (it needs to be overlapping with your A-B and A-C match).

Lastly keep in mind my other comment that even if you triangulate with B and C it's a very questionable segment when B and C are close family (parent/child or siblings). So you need distant (meaning cousins) relatives to form a proper triangulation when you only have two people to compare against.

---

Sorry Andreas that I upset you so much.  I still disagree with your statement that FTDNA cannot tell you where you and another person match.  Have you used the Chromosome Browser?  Have you downloaded and sorted the data from your matches?  Both of these will tell you both where you and another person match, and the size of the match, both in terms of cM and SNPs.  If two people both match you on the same location, with a sizeable enough shared match, then you can check the Matrix tool to see if they match each other.  If they don't, there is really no point in pursuing the question.  If they do, and neither of them responds, that may be all the answer you get.  As for the InCommonWith tool, that is a different issue -- it can be useful for some questions, but not many.

---

It is my understanding that in FTDNA's chromosome browser John Doe can see that he matches Adam from position 12,345,678 to 23,456,789 on chromosome 2.  John can also see that he matches Eva from position 11,234,567 to 22,345,678 on chromosome 2.  John can then check the Matrix tool and see that Adam matches Eva.

However John can match Adam on the chromosome 2 he got from his father and John can match Eva on the chromosome 2 he got from his mother and Adam can be matching Eva on chromosome 13 and Adam does not match Eva on chromosome 2.

I believe it is simply short sighted that Family Tree DNA does not yet offer real triangulation for autosomal DNA.  GEDmatch and 23andMe are the only ones I know of who do.

---

Well Peter explained it very well with his Adam & Eva example and segment positions!

It should be clear by now that you can't do triangulation like this without a critical missing piece! Again, this information is apparently withheld by FTDNA as they are worried about privacy.

It has nothing to do with being upset or not, it doesn't help when people (in general, not pointing fingers at you) repeat things over and over again that are clearly wrong. This leads to the believe that it's correct, just because so many people repeat it and reference to it.

The minority (and it unfortunately clearly is the minority due to the clever marketing of FTDNA to not tell their customer that they can't do triangulation with their tools) can only repeat and stress out what is correct many times over (even on this forum).

---

Thank you, Peter, for providing an explanation.  I haven't had this happen in my experience, but you show how it is possible in theory.  My question, I guess, would be how often does it happen in reality.  I can see that it would be a problem if both of your parents are related.  (Mine aren't -- I did run them through the GedMatch check for that.)  I would guess that it would also be a problem when dealing with endogamous populations.  But don't the problems with endogamy show up elsewhere?

---

Hello Gayel,

It can happen in reality if both parents are from the same "region".  "All" my mother's ancestry is from Georgia and earlier from North Carolina and the Colony of Virginia.  "All" my father's ancestry is from the Bahamas.   GEDmatch tells me my parents are unrelated.  There is at least one person who matches both my maternal uncle AND my paternal aunt.

Bahamians of European descent are endogamous, but I don't think of the Southern U.S. as being endogamous with the Bahamas (its those darn Loyalists;-)  I keep hearing about endogamy in colonial America. I can't wait for some smart person to create a endogamy calculator for WikiTree ;-)

Sincerely, Peter

---

I can actually see a match for someone from Georgia and the Bahamas as a "one off"   occurrence.  I have one match with my mother's cousin that is not shared by my mother, but my mother's cousin and I both have recent immigrants in our ancestry from the same area of northern Europe.  

I'm not sure how you would do an endogamy calculator.  Probably you would have to start just collecting the kinds of actual records of known endogamy, rather like what Blair Bettinger is doing with the actual ranges of segment match sizes for documented relationships.  His work really helps in moving us away from mathematical models into the realm of probability, which is where genetic genealogy really exists.  But, these are long term projects.

---

For an endogamy calculator in WikiTree I was thinking in terms of one's actual pedigree collapse (to what extent do you have the same ancestors on different ancestral branches in WikiTree?).  GEDmatch already has all the workings of an endogamy calculator (genetic) with its Are My Parents Related? feature.

---

That would be useful.

---

I have a real example from my own experience that illustrates how important it can be to have good detail on your matches.

23andMe and Gedmatch show that I have matches with 8 different people who are not known to me as relatives or connected to me by genealogy on the X chromosome in the area between about 117 million and 141 million (distances range from 13.7 cM to 33.2 cM). Since I am female, I have two X chromosomes, so these matching segments could come from either side of my family, and it's possible for an apparent match to combine segments from both of my X chromosomes.

When I compare the data for the other 8 people, I find that there are two distinct groups of matches here:

1. Sheryl, Gail, and Eric and I all match one another within that range.
2. Cari, Shay, Eowyn, H, Tammy, and I all match one another within that range -- but I am the only person in this group who has a match with the other group.

If I hadn't been able to compare the other people's data in detail, I might not have been able to figure out that these are two distinct groups of DNA cousins -- we are not one biig family.

Comparisons of the individual pairs also reveals that the boundaries of the matching regions are not very "clean" (my word, not a DNA term). What I mean by that is (for example) that Sheryl and I match from 118.5 million to 141.3 million, and Gail and I match from 122.4 million to 139.million, but Sheryl matches Gail from 120 to139 million. Since my matches with Sheryl and Gail include the territory between 120 million on a 122.4 million, when I see that Sheryl and Gail match in that area but that matches with me don't extend below 122.4, I conclude that my apparent matches with those other women include some material that isn't part of an "identical by descent" DNA segment. That sort of oddity underlines the potential for mistakes in DNA matching.

---

PS - If only there was some genealogy to identify Most Recent Common Ancestors with these two groups, I might be able to confirm the relationship by triangulating. I believe (based on other DNA considerations) that Eric and I are related on my maternal side, so  I figure that the common ancestor for group 1 would be on my maternal side. Since my X-DNA that matches to the other group must have come from my other parent, I conclude that group 2 must be related to me on my father's side (specifically, on my paternal grandmother's line, which is where my father's X chromosome came from).

Answer 1

To triangulate DNA, you need to have DNA matches -- three people who trace their ancestry to a common ancestor AND who all match each other on the same DNA segment. It usually takes a lot more than 3 people with tests to find a 3-way match to an 18th-century ancestor. And i don't think you and your mother really count as different people in that context, since you share half of your autosomal DNA with her, and you're probably  needing to confirm a match on a pretty short DNA segment.

Carrie has given instructions on how to look for a DNA match. May luck be with you!

Comments

Ellen is spot on. I just found a case where my girlfriend has 3 triangulated groups at the same location (which obviously is wrong). Turns out one of them is just a triangulation with a parent/child combination of kits.

So please take these sorts of TG's very carefully, I do understand that we all have locations with very little matches but it needs indeed 2 other distant relatives (distant from yourself and between each other) to form a proper triangulated group!

Obviously the more people you have in the TG the better (more family trees, more robust). If you can find the MRCA and CA couple with only 3 different family trees then consider yourself very lucky. Keep in mind that you need 3 different branches of their family trees as people might match through several branches of their family tree (through intermarriages/cousin marriages).

---

To add further to it, I've found 4 "fake" TG's now that only consisted of a parent/child kit combination for my girlfriends DNA test result. All of them on chromosome 1 which greatly reduced her overall number of TG from 6 down to only 2!

Three of these "fake" TG's didn't even match with any of the other matches at the same locus, further indicating that they are not a valid triangulation.

So please be aware and use parent/child TG's only when they match other kits as well. Then they are an excellent starting point as you now know which site of their family tree the MRCA is on.

Answer 2

Hey Michele!

Sounds interesting.

OK - so, fill in a few details here for me please.  Do you already know how you descend from Matthew Patten?  For me, knowing my relationship to an ancestor is critical when trying to use DNA evidence in support of a relationship.

Here is an example from my research.  Long story short, I am trying to determine if my gr-gr-gr-grandfather, John J. Coates (1815-1867) is related to Jeremiah Coates (1817-1896).  John lived in Meigs County, Ohio, while Jeremiah lived in Gallia County, Ohio.  Meigs and Gallia are adjacent on the map.

I don't have any paper documentation that strongly supports their relationship, but looking at the information from census records, etc. makes a relationship (perhaps they were siblings - or perhaps first cousins?) seem plausible.  Fortunately, there is some interesting DNA evidence to consider.

I have 8 tests (on Ancestry DNA that have been uploaded to Gedmatch.com) from descendants of John J. Coates who match one another (the 8 tests are first cousins, second cousins, and second cousins - 1 X removed to one another) and 2 tests (Ancestry DNA uploaded to Gedmatch) from descendants of Jeremiah Coates (the two testers are third cousins).  Just looking at the respective family trees of the 8 tests, John's descendants would be 4th cousins of Jeremiah's descendants if John and Jeremiah were siblings (and 5th cousins if John and Jeremiah were first cousins).

Similar to what you can do on FTDNA, I have run one-to-one comparisons between each of the 8 tests from John's descendants to each of the 2 tests from Jeremiah's descendants.  From this, I see that 5 of the 8 tests from John's descendants match Jeremiah's descendant, RKH, and 5 of the 8 tests from John's descendants match Jeremiah's descendant, Duane.  Looking at these basic comparisons, I can identify 10 different DNA segments that exist between a descendant(s) of John and a descendant(s) of Jeremiah.

Taking this to the next level, I then ran a 3D chromosome browser comparison for the 10 tests.  From this, you then see that, in 4 of the 10 DNA segments, two of the descendants of John share the same segment with one of the descendants of Jeremiah.  Arguably, these 4 segments are "triangulated" in that two separate descendants of John share the same DNA segment with one another AND with one of the descendants of Jeremiah (i.e. triangulated across three independent testers).  

I was hoping that the two descendants of Jeremiah would share segments in common with one or more of the descendants of John, as this would be very strong evidence of a relationship between John and Jeremiah.  My current data did not produce this result.  But, since I only have 2 tests from Jeremiah's descendants, my next step is to add additional tests from descendants of Jeremiah.

So - this is an example of triangulation.  I am using DNA evidence from multiple descendants of ancestors to explore a relevant genealogical focus.

Comments

Returning to Matthew Patten, it appears that you are trying to support the match you have found between your mother's test and the test from Ronnie (Jumper) Smith (RJS).  Since you are the direct descendant of your mother, we won't use your test, as your mother is closest to the ancestor of interest.

Since I don't know the relationship between your mother and RJS, given that Mathew was born in 1728 and had children in the 1750s-1760s, my guess is that your mother and RJS are 5th-6th cousins.  Distant cousins (4th cousins and beyond) typically only share a small amount of DNA (typically one segment between 10-30 cM in length in my experience), which makes it highly unlikely that a study including test results from just three 4th+ cousins will share a single specific 10-30 cM DNA segment.  As such, you will want to look at the DNA shared between multiple descendants of Matthew Patten to improve the odds of multiple tests sharing the same segment (triangulation).

So, going back to your current match with RJS, you should look at the match between your mother's test and RJS in the chromosome browser and note the specific location of your the DNA segment that is shared between your mother and RJS (for example - on chromosome 1 between 23,560,000 - 78,832,000 - which is 27.2 cM in length).  Then, if possible, you should upload your test results to Gedmatch (and encourage RJS to do the same) to see if you find matches with other testers who descend from Matthew Patten.  You can then start to note the respective segments that are shared among the various tests.  Any segments shared across multiple tests are evidence of triangulation.

The other critical point to remember, especially when you are looking at 5th-6th cousins, is that you have to account for the possibility that testers might have another common ancestor in a different part of their tree.  This is easier to account for when you are comparing first or second cousins, but when you get to 5th cousins and 6th cousins, we are talking about gr-gr-gr-gr-grandparents and gr-gr-gr-gr-gr-grandparents - i.e. a large number of ancestors/possible sources of DNA in each match's respective tree could be the source of the match.

---

Ray, my mother is Veronica "Ronnie" Jumper Smith.

Answer 3

I try to focus on looking for triangulated groups for matches who have a GEDmatch ID and their ancestry in WikiTree.   If they don't, then I invite them to join those two.  I also offer to help them join.  GEDmatch's triangulation utility and WikiTree's Relationship FInder are real time savers.

We will all benefit if you ask your matches in GEDmatch who have multiple kits to change the duplicates to Research or delete their duplicate kits.  A lot of AncestryDNA testers are uploading their same kit multiple times.

Comments

Peter, you identify those duplicate kits as they share more than 6000 cM between themselves?

I never come across any so far, though I have analyzed a lot. They could be identical twins, right? Not sure what the probability of identical twins is.

---

Many times, those duplicate kits in Gedmatch are identifiable not only by the shared DNA, but also by the fact that the people have the same name.

---

Not so sure on this Ellen. I've come across a lot of kits so far which have been named the same (sometimes only differ in being written with a capital letter at the start or not) and on closer look they turned out to be parent/child kits.

This still seems to be the most logical and most common combination in our hobby. As long as our parents are alive we should take the opportunity to talk to them about taking a DNA test.  If you get both to test then we can do phasing on GEDmatch which helps tremendously sorting out all the good vs the bad.

Answer 4

Thanks to all of you for your feedback!

Well, it seems as if this isn't going to be the quick little project I thought it would be. I have, however, reached out to a handful of 3rd to 5th cousins asking them to consider testing and giving a very lay-person explanation of what we can expect to either confirm or disprove.  Now, I wait to see who gets results back and how long it will take.