A Segment-ology TIDBIT
Start with a list of your overlapping shared segments – they all match you (two triangle legs). Find two of these shared segments that match each other (third triangle leg) to form a Triangulated Group! All the rest of the shared segments in your overlapping segment list need only match one of the shared segments in the TG – any one of them with a good overlap – to be added to the TG. The explanation will take a long blog post with diagrams – but the thrust is that forming a TG basically identifies that segment just as good as trio-phasing. So trust me! If a shared segment doesn’t match the TG, it will match the other, overlapping TG; or it is IBC (it happens for some shared segments under 15cM). As a Quality Control measure, I often make a second comparison – it always matches!
[22D] Segment-ology: Only One Comparison Needed to Add a Segment to a TG TIDBIT by Jim Bartlett 20170101
segment-ology 
Jim, that’s great, I think I finally got it straight. I am going through some of these stacks now. It does seem get a little weird transitioning from one orderly column to the next, with a couple of matches sort of straddling both. I think I just need to go through them all and see how they wind up. Thank you so much for the guidance and support. -Matt
LikeLike
Matt
You are welcome. Some of the larger segments will straddle more than one TG. In general they are closer cousins. Here’s a way to visualize straddling: you got a full Chr 01 “segment” from your father – it straddles a lot of TGs. Even when you break Chr 01 down into 4 or 5 (typically) large segments from your two paternal grandparents – each on will usually straddle several TGs. Etc.
You might want to re-read the Bottom-Up and Top-Down blog posts with this concept in mind.
Jim Bartlett – atDNA blog: http://www.segmentology.org
>
LikeLike
Hi Jim,
this sentence isn’t 100% correct based on my experience: “All the rest of the shared segments in your overlapping segment list need only match one of the shared segments in the TG – any one of them with a good overlap – to be added to the TG”
That one person matching could be you, meaning that the added person matches you on the other chromosome pair. Would still fulfil your condition of just one match needed, right?
Happened in my data so the case isn’t so rare that it’s not happening. I always check for two matches in the TG just to be 100% sure.
LikeLike
Andreas – I think it is 100% correct. The context is “your overlapping segment list”. Any one of the your overlapping segments, which match some other overlapping segment in “your overlapping segment list” will Triangulate. What I’m saying is equivalent to the initial Triangulation. To be fair, I almost always test against at least two shared segments – just out of caution. But I’ve not found any that don’t match. Perhaps the difference here is that I have thousands of Triangulated segments in (currently) 361 TGs. For most of the new shared segment adds, I can predict which TG (paternal or maternal) the new one will fall in. If it’s in a sea of blue segments, it’s going to be blue. There are only a few places where the start locations of shared segments (within TGs) overlap each other and a guess is not so clear cut. However, with this much “coverage” of established TGs, I only need to check one side about half the time (and a follow up with another segment for Quality Control); the other half the time I check one on the other side to get a Match (keeping an eye out to insure a good overlap). Only a small percentage of the time will a shared segment not match either side – I usually check these against several on TGs on both sides before I classify the segment as IBC and remove it to a different spreadsheet. We need care at this point, because sometimes there is no match in the TGs because there is not sufficient overlap – I code these “Undetermined” until more overlapping shared segments can be found.
LikeLike
Jim, I think I am finally able to wrap my brain around the triangulation process. I appreciate all the time you are expending answering our newbie questions, so thank you! I have a couple more… 🙂
If I may try to oversimplify the process talked about in this post, I might see where I am having trouble reconciling all the concepts. Here is how I am understanding it. Lets say I have a list of ten matches to my DNA on Chr 01 that all overlap within a given segment that is well above the valid Cm threshold in length. If I check the first two matches against each other and myself (Me-1, Me-2, 1-2), and they all match each other, I have a TG consisting of matches 1 and 2 (which of course also includes my segment), lets call it TG-1A. And the definition of a TG says that all three segments must be on the same Chr, either M or P (we don’t know which yet). If I understand what you are saying in this blog post, the other eight segment matches in my list can then be included in the same TG-1A if they match against either one of the original two matches that are already members of TG-1A. These are just additional triangulations to the same Chr segment. Am I good so far?
OK, here is my gap in understanding, and maybe it is just because I have little experience with all of this. If I have my first two matches in a TG, and the other eight are stacked up in a nice column (if I am looking at them in a chart), with their start and end points lining up nicely, is it probable that they are all in the same TG, or is it you just can never tell and have to test each and every one against one of the “known-good” members of the established TG for that segment range to verify it is indeed on the same Chr as the TG? Writing this out I might have answered my own question. I guess I was looking at it as if it seems like that if one of the ten match segments is from the other Chr, how is it possible (probable!) that it would line up with all the others in the TG so nicely? But I guess it could just be coincidence, and the one segment could just be from the opposite Chr in exactly the same segment region, but contain a totally different DNA sequence HIR match. Yes? Good? I need a beer, my head hurts a little…
Thanks, Matt H
LikeLike
Matt
You are good so far! The reason why you can compare with just one other Match (well technically, with one other segment) is because you and all these shared segments are already 2 sides of a bunch of triangles – so when you compare a new shared segment with anyone in the TG, you are testing the third leg. Now, about stacked segments 1. Unless they are all over 15cM, you need to test each one to insure it’s IBD. 2. With a lot of experience, you start to get the feel of whether a stacked group are going to be pretty much all one TG, or whether it’s a salt-n-pepper mixture – I’ve seen them both. However, usually, most stacked segments will go into one TG. But that’s a calculated guess, and certainly not a guarantee. Tell you what: try a bunch one at a time, and let me know. I play a game with new Match-segments: I try to guess which of two overlapping TGs they will “join”. I’m right about 90% of the time – not good enough for me, so I triangulate each one.
Jim Bartlett – atDNA blog: http://www.segmentology.org
>
LikeLike
I have 5 individuals who align up on the same segment area on the same chromosome number (say 8) in GEDmatch chromosome segment matching. However, using GEDmatch one-to-one compare, individual D does not match A or E at that segment area. So that gives me triangulation groups ABC, ABE, ACE, BCD, and BCE. Does that mean that individuals A, B, C, and D match on one of the chromosomes 8 as one big triangulation group and individuals B, C, and D match as one triangulation group on the other chromosome 8?
LikeLike
It looks to me like all 5 are in one TG, and D probably just doesn’t overlap A or E enough? Lower the thresholds a little. Shared segments cannot be on both chromosomes, except special cases like siblings.
Jim – http://www.segmentology.org
>
LikeLike
Lowering the thresholds showed a match between D and A but even with a threshold of 1 there was no showing of a match between D and E.
LikeLike
Dennis, How many cMs in the shared segments: AD and AE? This does not make sense to me.
LikeLike
I read your work with great interest.
Perhaps I am missing something – but as I understand it, in the case of a double crossover, two short(er) segments are created – one being maternal and on being paternal – both being the same size.
If subsequent mitosis events occur in subsequent generations and no crossovers occur to disturb the created segments, many people could have the same segments, half of them from the original maternal parent and half from the original paternal parent. And they would be at the same place on the particular chromosome. So is there not a 50% chance that a segment of that size at that location would be (relative to that ancestor) maternal in origin and would not match those which were paternal? I realize that several conditions must be satisfied for this to happen, but can we really say that a segment as described comes from only one ancestor when there are other segments which came from another ancestor (a spouse?).
Thank you,
Vance Wiley
LikeLike
Vance,
There are several extremely unlikely scenarios that could complicate the issue. There is a chance that segments that are separated could be stitched back together, for instance. In general, each child gets a different, random jigsaw puzzle of segment pieces from independently created sperm and ova. And even if they did have the same, or nearly same, ancestral segment, they would have to pass it down, intact, to descendants who, as cousins, eventually marry. In this case one is a male and one is a female – so it would be possible for a child to get the same ancestral segment on both the maternal and paternal chromosomes (again very unlikely). This would exist for this one generation only. In the next generation, the child can only pass one of those segments on. So in succeeding generations, there would again only be the one ancestral segment. So, at a very remote possibility, two people could pass the same ancestral segment to their child (one on each chromosome), but the child can not pass them both on. You might want to review my blog posts on endogamy. Another way to look at this issue is to understand that you only get paternal segments from your father. He got them from his mother OR father (he can only pass down one or the other). Whoever it was, likewise got that DNA from their mother OR father. Repeat going back as far as you want. Each segment of DNA came from one parent, one grandparent, one Great grandparent, etc. Again, read over my endogamy posts. Jim
LikeLike
Vance, I should add that a double crossover does not create a duplicate segment in a child. There is still only one resulting chromosome that is passed from the parent to the child. The other parent is not involved in this double crossover.
LikeLike
What a great way to start the new year with five new posts. Thanks, Jim.
LikeLike
Dennis, you are welcome – hope they help. Jim
LikeLike
Love the tidbits! Thanks, Jim.
LikeLike
Thanks, Nancy – it’s easier for me to get little points out there.
LikeLike