Thanks to all for your responses to my last blogpost. All of them are a good read.
I had always thought a TG segment was crystal clear… WRONG. Per the classic refrain from the Legal Genealogist, Judy Russell: “It depends!” My second ever blogpost on 9 May 2015 (Benefits of Triangulation) stated 16 benefits, including: Organizing most Matches into TGs; All Matches in a TG have the same Common Ancestor; the TGs define crossovers and a Chromosome Map; TGs are equivalent to Phased data. What I didn’t say explicitly is that each TG represents a segment of my DNA.
The elephant in the room is: who was the first Ancestor to pass down that segment (as part of a full chromosome passed to a child who is my Ancestor)? In other words, in what earliest generation did that full segment first exist in my line? There may be a different such “elephant” for each Match… but that’s another story.
So back to “it depends”…. For me there are 3 objectives:
1. “See” my DNA segments. Divide up my chromosomes into discrete segments, each one of which came from a specific Ancestor.
2. Determine the Ancestor for each segment.
3. Determine my Chromosome Map of segments – each segment being adjacent to another segment from the beginning to the end of each of my 45 chromosomes.
When I started forming Triangulated Groups, I only worked with known cousin Matches. It created a patchwork of TGs. One day I decided to bite the bullet and Triangulate all of my segments, a company at a time (FTDNA, 23andMe and MyHeritage). It took months without many of the tools we have today. And the three versions meshed virtually exactly! That was as expected since all comparisons were against my DNA. I was using the “full” version of a TG, plus some judgment for large segments from close relatives that spanned more than one TG. This brings me to a significant factor in Triangulation: Judgment.
Judgment: It’s easy to compare yourself to another Match and “see” an exact shared DNA segment. But what would happen if Match 3 in the last blogpost only overlapped Match 1 by 5cM? Would we then call this a 5cM TG (against the rules and throw the whole thing out?). Would we discard Match 3 (even if they had a robust Tree that included a CA)?
Judgment: Sometimes there is a close relative, Match 5, who overlaps much more than me and Matches 2 and 4. Experience (and judgment) tells me that this somewhat larger segment is probably from a close relative whose Common Ancestor with me includes a father/mother more distant – with one of them being the CA for the full TG.
As I read over the comments of the previous blogpost, several words pop into my mind: context, messy, complex, judgment, imprecise, etc., as well as “we’re making this up as we go”.
Messy – yes Triangulating all of our Match segments against our own can be messy – and judgment is needed. Given the random nature of recombination, I do see some curve balls from time to time. Triangulation usually identifies false (IBS) segments, which should be discarded. If I find a shared segment that really messes things up, I’ll also discard it (or at least highlight it as weird). As I’ve blogged before, the raw data is sometimes messy – or fuzzy – sometimes reporting a shared DNA segment that runs longer that it should. Although my parents are not related (per GedMatch), I do have one area of my DNA that my two parents combined have all of the most common SNPs, and so I get a “zigzag” pileup of many Matches with false segments there. I’ve identified this area and then toss out those Match segments (<10cM). Pedigree collapse and endogamy also create messy areas. To the extent possible, identify these specific locations with a dummy segment to highlight the potential issue.
Context – in developing my Chromosome Map, the segments will be adjacent to each other. I look for the previous and the following TGs to the one I am working on. Ideally (and actually) each of my segments will “crossover” to the next segment which is from a different Ancestor of mine. Note – that “next” Ancestor may involve a different grandparent, or a different 3xG grandparent. We have to fill out the Chromosome map to figure that out, but it is important to remember that the next TG will have a different CA. So if I accept the conservative TG (a part of the Match 3 shared segment), what different Ancestor can I find for all of the “leftover” shared DNA segment pieces of my DNA.
Complex – One complex part of this analysis is what about the parts of true segments from Match 1 and 2 and 4 that are not in the full TG I show in blue? I focus on my DNA, but I think every true Segmentologist should try this experiment with say Match 2 at GEDmatch. Use Segment Search to find other Matches who share the same segment and build the TG for Match 2 – it “will” be different than my (or your) TG. A little different or a lot different? If Match 2 is a known cousin, the same MRCA would almost always apply. By doing this with other Matches in a TG, many of us (working together) are building a larger segment of the CA.
Imprecise – I’ve blogged about fuzzy data. I counter this with judgment. I look at all the segment data for a TG (all my segments are in one spreadsheet). Among the TG fuzzy start data I decide on a specific Mbp start location. Then I decide on a Mbp start location for the next (adjacent) TG. Often some shared segments from the initial TG will “spill over”, past the start of the next TG. The small amounts of spillover, I just ignore: fuzzy data. If there is a large spillover, I’ll consider if the second TG is potentially closely related to the first TG, or not.
Imprecise – This also describes the fact that all your shared DNA segments may not “cover” all of your DNA neatly, or uniformly, or even completely. The shared DNA segments are independent and random – they are not at our beck and call… They don’t necessarily help us fill the gaps perfectly. They are what they are – they are clues we must use as best we can.
All of the above is to indicate that all IBD shared segments should have a home in a TG, and that all the TG segments should cover all of your chromosomes, IMO. Remember, at each generation, all of your segments from that generation must add up to all your chromosomes!
Another aspect of this which I muse about is the SNPs – thousands of them in a unique arrangement in my DNA. Let’s say Match 1 shares 2,000 SNPs with me. Alone we would say the shared DNA segment between us (green) came from a Common Ancestor. Similarly we would say the 3,000 SNPs in the shared segment with Match 2 was from a CA. I don’t see how we could argue that these two CAs were somehow different. I think it is much more likely that the CA is the same, and Match 1 just didn’t get the full segment that I did and Match 2 did. Match 3 is in the middle of all these SNPs – surely Match 3 got the same SNPs from the overlapping locations. By comparing the SNP values of all 4 Matches, I’m confident that we’d find the same values at each SNP location.
Note: all of these Matches and evaluations are based on separated cousins. Of course close relatives could have the same segments and SNPs – the whole concept of segment Triangulation depends on an analysis of more distant relationships.
My summary:
The TG Group of Matches should all look for the same Common Ancestor – and hopefully help each other toward that goal.
The full TG segment (blue) is my DNA segment, which I can use as part of my Chromosome Map. It defines my crossover points. Also I can contribute my SNPs to any larger study of my Ancestor’s DNA.
I must be careful to not state that my Matches have this TG segment. Matches will have their own different, but overlapping, TG segment.
The Common Ancestor almost certainly passed down a larger DNA segment, through at least some of their children, which different descendants (including some of my Matches) got. Note: there may be other descendants who have DNA tested who may share with the TG Matches, but not me (I am not the center of the universe…)
[08Ab] Segment-ology: Boundaries of a Triangulated Segment Part 2 by Jim Bartlett 20250915
segment-ology 
Thanks, Jim, for the clarification. I am using DNA Painter’s handy-dandy chromosome mapping tool, where I have mapped 545 segments (heavily skewed to my paternal side), but haven’t generated statistics on the number of TGs or ancestors they represent, which of course “telescope” back through multiple generations in some cases. And yes, the fuzzy segment ends have been noted for some time, although using GEDmatch’s “Segment Search” tool, I am surprised how many segments from different matches do share the exact same start or stop locations, even when uploaded form different testing companies. This tool is a kind of automatic TG generator (with caveats).
LikeLike
Todd, yes the DNA Painter tool is a great go-to tool for folks who prefer visually “see” the shared DNA segments. Each of these shared DNA segments will almost always be only part of the full segment you got from the Common Ancestor. What remains to be done at DNA Painter (and likewise in my spreadsheet version of exactly the same data) is to tease out the crossover points in your own DNA and define the full segment you got from an Ancestor. These grouped segments (either Painted or data) are what makes up your Chromosome Map Each of your Matches would almost certainly find a different “grouped” segment from the Common Ancestor (which overlaps your TG segment).
Your idea of GEDmatch as an independent arbitor for crossover points is a good one (albiet for only a fraction of Matches we have)
Your “telescope” analogy is also great. Going back – from you to an Ancestor, a TG segment will only be part of a larger segment (like a telescope) until it’s part of a full chromosome in the Ancestor who first had it. From the Ancestor down persepective, it starts as a full chromosome passed to a child (with the imbedded TG segment), and in each succeeding generation coming down, the segment (with the imbedded TG) gets smaller (or stays the same – sticky), until it gets to you. Or a full chromosome is passed down to the Match with the Match’s TG segment imbedded in it at each generation. The only thing we know for sure is that the shared DNA segment between you and that Match is part of every segment passed down through both paths to you and the Match.
Thanks again for your insights, Jim
LikeLike
This is partly a test message to see if my WP account is properly set up… But my question may be addressed under “Context” above. I was wondering if you can define where the TG begins to the right of the TG in your illustration. Does the next TG have a defined starting line, and is it to the right of the longest green segment? I’m not sure what segments you are working with. Are these segments unknown relationships? I only add segments from known (or reasonably projected) relationships to my map so that I know what I’m dealing with. Often there is a cluster of matches belonging to a TG who share a surname in their trees, but the precise genealogical connection is unknown. These are “surname triangulations” with unknown CAs. So I have defined “windows” (using my analogy) except for segments belonging to closer relatives who would descend from adjacent TGs, so those segments randomly cross over the defined lines between more distant ancestor TGs.
LikeLike
Todd – you spotlight the conundrum elephant in the TG room. But first to answer your questions: The TG to the right of the illustrated green TG starts with the next base pair on that chromosome. We don’t know that, so it’s the next SNP; and the precise SNP is also fuzzy. So the squishy answer is each TG begins immediately after the one to the left ends. The segments I’m working with are all shared DNA segments with different Matches. Shared DNA segments are basically all we’ve got to identify parts of the true segments we get from our Ancestors. In one trivial sense I “know” all of my segments – they are 22 chromosomes from my dad and 23 chromosomes from my mom. Most of these are subdivided into fairly large segments from my four grandparents – and each of these start where the others end – they must be adjacent to each other and account for every base pair.
The TG conundrum is caused by two different factors: 1. In which generation was the TG first formed; and 2. In which generation is each Match (are they a 1C, 3C1R, Half 5C3R…?). The TG conundrum is that the TGs we identify as being side by side may not be all of the same generation. And, if we lower our cM threshold enough, every TG segment will subdivide into more distant TGs. Because the DNA is random, we cannot pick a cM threshold that will guarantee all of our TGs are based on only, say, 5xG grandparent segments… Maybe using the LEEDS criteria we could come close with grandparents. Forming TGs is mechanical, and with a little genealogy, I’ve been able to identify all 372 of my TGs as paternal or maternal. So I can “paint” them all – even though I don’t know the correct grandparent in a few of them. I have surname/ancestor triangulation with multiple DNA Matches at Ancestry out to all of my known 5xG grandparents. But I cannot connect them to TGs. Actually with 64 different 5xG couples, and 372 TGs, I have about 6 TGs for each couple. I’d need to go back two more generations to get about 1 TG per Ancestor couple. So, I have some TGs with unknown grandparent Ancestors; and I have a number of Match Clusters that I cannot link to any DNA segments; and I have several “Floating Branch” Clusters of Matches with no known TG or Ancestor. Lots still to figure out.
Hope this helps,
Jim
LikeLike
So it sounds like you’ve uploaded all segments from all matches into your map, including unknown relationships and close relationships? Have you tried mapping only segments shared with matches who have confirmed, more distant relationships? I include only those segments that can be identified as inherited from an individual great-grandparent (I don’t use couples) or beyond. My chromosome map in DNA Painter has almost full coverage for my paternal side (Southern colonial) and very sparse coverage on my maternal side (German and NY/New England). So I have fairly discrete TGs and don’t see the messiness/fuzziness you describe with lots of overlap. But we may have different objectives in mind. My particular approach to forming TGs may be obscuring my understanding of what you are describing. Sorry if this only confuses matters!
LikeLike
Todd – Yes – at one point I had all of my shared DNA segments (over 7cM) from FTDNA, 23andMe, MyHeritage, and some at GEDmatch, in one spreadsheet. Of course, I formed Triangulations within each company, but because my DNA was the base, they all came out almost perfectly in alignment (I had to shave a few MH segments). All of these shared segments formed TG segments that covered my 45 chromsomes, except 4 or 5 small areas where I just declared a segment to fill the void. I think I’m down to only 2 such filler TGs now. I have close relatives that agree with those areas (as part of larger segments), but no cousins do. Anyway, that is one definition of a map – those crossovers in my DNA are now fixed. It’s like an upside down jigsaw puzzle. Out of 372 TGs I have 49 specific TGs no Matches with MRCAs beyong my grandparents. I have many TGs with MRCAs out to 4C to 8C – who knows, at this point, if that’s as far back as I can go on them.
Re lots of overlap: some of my TG ends are pretty clean, some are fuzzy – by fuzzy, I mean tips of TGs overlapping 1-3 Mbp – not significant amounts of overlap, but many cases of small overlaps. Probably due mainly with how the different companies define shared DNA segments.
Like you I have better coverage for 3/4 of my Ancestors going back to Colonial VA, and less coverage for my Scotland-Germany couple that immigrated in the 1860s…
It looks like we have fairly similar outcomes.
Jim
LikeLike
Todd, Let me also add, that I haven’t updated my TG spreadsheet in several years – I put it away (because it was so hard to find CAs at FTDNA, 23andMe, My and GEDmatch) and shifted to Ancestry/ThruLines/ProTools to find the Matches with CAs. I now have multiple confirming Matches who are cousins on all of my known Ancestors back to 8C level – I have 8,000 rows of Matches and MRCAs in that spreadsheet. Oh, how I’d like to have a browser there for a week or two and confirm the TGs for those… J
LikeLike
I think of a TG as a “window” on your common ancestor. You know that objects seen through the window extend beyond your view, and some do not span the full length of your view. You are limited by the dimensions of your own window. However, with what you can see from your own window, you can work back to the fuller picture of a common ancestor if you have enough information.
When 23andMe offered a chromosome browser, I could easily compare shared segments against our chromosome paintings. I would compare small segments of Subsaharan and Native American DNA with matches who shared segments in those locations to see whether they matched the ethnicity there or not. And if they did, it was sometimes immediately apparent that their “ethnicity segment” was longer than mine, extending beyond the view of my “window”, or shorter, when my segment extended beyond their view.
I’m currently working on a long segment shared with several matches that is comprised of a smaller Subsaharan segment at the left end, and a longer European one extending to the right end. Some matches (who are confirmed 5th cousins, and assumed 6th cousins) share the entire length of the segment (full window) which includes both Subsaharan and European segments, while other matches share only the Subsaharan segment or only the European one. These latter matches, I believe, are more distantly related than the ones who share the whole segment, and I have good clues from traditional genealogy how they would connect to my tree.
LikeLike
Todd Thanks for your feedback – a great analogy. So it looks like your long segment was from a 4xG or 5xG grandparent, and probably split to his/her grandparents going back.
Jim
LikeLike
Completely agree with you Jim. And then I go to my chromosome map and discover that yes, I appear to have a TG at one location, connected to just one family line. But then I have another segment from some other lab that appears to overlap part of this TG for a reasonable distance – let’s say 10cM.
And this segment links to a rather sparse tree – certainly not back the 8 generations I would like, probably only two or three.
At present my working hypothesis needs to include the possibility that, way back, there is DNA from someone in this line who has married into a totally different line of mine.
Now I do need to go back and see how many SNPs there are in that overlap region to check that the overlap is sound.
Thus I often find a difference between theory and practice, and am slowly beginning to understand why that might be.
LikeLike