Triangulation at 23andMe

A Segment-ology TIDBIT

23andMe has a great feature. It starts out as a standard In Common With (ICW) list for each Match you are sharing with (old Shared Genomes Matches and new Open Sharing Matches). This ICW list is near the bottom of your Match’s page. But the difference with other ICW lists is the Shared DNA column. The Matches marked with a “Yes” have overlapping segments – and over 99% of the time they form a Triangulated Group (TG).

So go to your DNA Relatives page and scroll to the bottom and click on the “Download aggregate data” link. You’ll get a spreadsheet of all your matches and most of the 23andMe data. Sort the spreadsheet, and delete the ones with no segment data. Then sort on Chromosome Number and Chromosome Start to put them in a particular order. Add a column called “TG ID”. Now you’re all set to begin Triangulating.

Start with the first Match in the spreadsheet (let’s call it A). Click on the hyperlink* that takes you to A’s page, scroll down to the ICW list and note in your spreadsheet Match A and each Match with a “Yes”. Since you are starting on Chr 01, call this TG: 01A, and put 01A in the TG ID column for A and each “Yes” Match. This pretty much identifies all the other 23andMe Matches that are in a TG with (A). The whole TG 01A (of 23andMe Matches) is created through one Match! There may be a few that don’t overlap A enough to form a shared segment at 23andMe, but all you have to do is go down your spreadsheet list of 23andMe Matches and select the next Match (B) that is not already in a TG; click on B’s hyperlink and look at their ICW list for “Yes” Matches with (B) – some will either overlap with (A) (call all of them 01A, too); or they all form a new TG (say 01B) – one or the other. Then continue with the next Match not already in a TG.

One could probably go through their entire 23andMe list of shared Matches in a few hours, creating TGs for all of them. There may be some with no ICW “Yes” Matches – give them their own TG; and move on. Be careful with Matches with more than one shared segment – make sure to treat each segment individually – this may take a little extra analysis.

Remember, TGs represent segments (from an ancestor) on one of your chromosomes. They are equivalent to phased data. I consider all shared segments at 23andMe which Triangulate to be IBD. All of them should be in a TG on one side (parent), or the other.

If you have known relatives in any of the TGs you can assign those TGs as Paternal or Maternal. This often allows other, overlapping TGs to also be assigned to a side, using logic.

*Note1: clicking on the 23andMe hyperlink (Link to Compare View) is a little tricky – I usually just copy (Cntr-C) the spreadsheet URL, and paste (Cntr-V) it into the URL bar of any open 23andMe page – hit Enter. It goes pretty fast.

Note2: feel free to use any TG ID numbering system you want. I think it’s wise to start with the Chromosome number. But you can name your TGs Bill, or Bob, or Sue if you want. You are creating groups that will tie to ancestral lines.

ARE YOU READING THIS FTDNA? ALL YOU NEED TO ADD IS A YES!! And AncestryDNA could add a similar feature, and hell might freeze over, too.

Enjoy easy Triangulation at 23andMe…

 

[22M] Segment-ology: Triangulation at 23andMe TIDBIT; by Jim Bartlett 20170720

54 thoughts on “Triangulation at 23andMe

  1. Thank you for this tip!! 23andme just went up in my view of DNA sites. I agree with your comment about Ancestry. I don’t understand their reticence to give the serious DNA researcher help. All of the other major sites have some sort of Chromosome browser and give you access to the CH data. If I found a site with as good support for building a traditional tree I would switch in a heartbeat.
    Doug

    Liked by 1 person

  2. Hi Jim. I am in the process of organizing chromosome data for my mother and I using downloaded chromosome matching data – it seems to be a lot of work to establish TGs due to so many thousands of matches. I also have both of us tested at 23andMe and so will do as you suggest above. My question is, can I merge the two into one set of TGs and one chromosome map for each myself and my mother? Are the segment addresses (start and stop points on each chromosome) transferrable between the two sets of data? It just seems like a real bear of a job to sort this data to the point where it is usable for genealogy! Do you suggest keeping the lists separate or joining them?

    Like

    • Douglas – the quick answer to your final question is – both!
      Each of your parents has many recombination crossovers in their DNA – in both of their own sets of chromosomes. They pass roughly half of the DNA in each of their chromosome sets (1-22) on to you, AND they create new crossovers points in the process. So no to the idea that somehow their segments have the same Start and End locations as yours. You must form a separate spreadsheet, TGs and chromosome map for each. My strong recommendation is that you have a field (column) in each spreadsheet for the test takers name, or initials. Then you can combine the two spreadsheets to make some comparisons and analysis; but then you can separate them. In practice I keep them apart almost all the time; and combine them every 3-6 months for a few days of analysis; then separate them again.

      Liked by 2 people

      • Jim sorry but you misunderstood my question. My question is along the lines of – is the chromosome location/address data between FTDNA and 23andMe compatible… ie using the same addressing?

        Like

      • Yes. They use the same standards. At gedmatch people who have uploaded from different companies all have shared segments with me that are essentially the same.

        Like

  3. Thanks for extolling the triangulation tool at 23. I have only been using all 4 testing companies for barely a year and clearly find myself checking into 23 everyday because it seems I have more contacts there. The triangulation tool is a winner ( my other daily go-to is GEDmatch). I like that I can post my public tree link at 23 as well as the messaging system. With the other two companies, I need an aspirin to work on them. I love FTDNA but sure wish they did this kind of triangulation within the matrix or ICW. Thanks for the spreadsheets tips, Jim. Lifesaver as usual.

    Liked by 1 person

  4. A lot of the relatives in common seem to have “Share to see” instead of a “Yes”. Would you include them as if it said “Yes”, or do I have to ask them to share and wait until they do?

    Like

    • No. Yes. Look at the ratio of Yes to the total list – usually a small percentage. The Share to See Matches are also probably not Yes. Try it and see how many Yes’s you get. I will often click the Share to See for my largest shared segments – not for the “Yes” here, but for the several segments and “Yes”s wherever they are.

      Like

  5. Jim this is wonderful to know and a great feature.

    I recently was shown the new circles at Ancestry program with the addition of DNA instead of only tree comparisons. Though they still don’t show the chromosomes, by moving the cursor around you can see who shares DNA that links to others. It is an improvement.

    Like

    • Triangulation is a mechanical method to group your shared segments – each group has a Common Ancestor. But determining the Common Ancestor is a genealogy process – you need to share info with your Matches, review their Trees, communicate with them, often do some research, etc. to determine the ancestor(s) you have in common.

      Like

  6. Jim, do you have a quick way to add just the new matches at 23andMe to an ongoing spreadsheet? The only way to get their full info is by downloading the aggregate data, isn’t it? I have downloaded the data a number of times from 23andMe and from FTDNA, and get bogged down trying to sort it out and keep it straight, especially since the two companies format their csv columns differently.

    Like

    • No; Yes. It is a PITA. I rearrange the columns in the download to match my master. Then add them in and sort on name and delete the duplicates. Like I said – a PITA. Some use Genome Mate Pro; some us DNAGedcom Client to do the downloads and updates.

      Liked by 1 person

      • Jim, what is the advantage of triangulating using this tool vs doing it by downloading the data and sorting it in a spreadsheet? Seems like I get the same results in less time with a spreadsheet. Am I missing something here?

        Like

      • mizmdk – On the one hand I can download the spreadsheet of all 23andMe Match-segments and put them into my Master spreadsheet. And easy to sort the spreadsheet in Chr and Start to put overlapping segments next to each other. But there is still two problems: which TG do they match, and where do TGs Start and End. Using this 23andMe feature solves both of these.
        In my case I’m routinely adding new Match-segments (and the Match hyperlink) into my spreadsheet. So for each new segment, I just link to the Match page and check a few “Yes” Matches against my spreadsheet to determine which TG the new Match is in. It goes pretty fast.
        Any method you use, should give the same results. All of the Match-segments “fit” in only one way.

        Liked by 1 person

      • OK, thanks for clarifying this. Where to start and end each segment has been difficult for me to determine, as there is so much overlap in both directions when you look at the spreadsheet.

        Like

  7. If only these people were interested in genealogy in any way. I’ve been adding a few here and there to my triangulation groups that I make at ftDNA and GEDMatch. If I can find one with a GEDMatch id who overlaps with a known member I can get the whole group mapped to side. But 20 people triangulated without a tree or interest in genealogy is pretty much just frustrating.

    Like

    • I agree, and wish more had Trees and interest. But the other side of the coin is if we limited test takers to only active genealogists with robust Trees, we wouldn’t have many Matches or TGs. At least all these kits let us create many TGs which helps define the boundaries of segments from our ancestors – it’s an important grouping mechanism. And, with all the interest, more new close Matches are coming in. More than enough fodder to keep me busy.

      Liked by 1 person

  8. Thank you so much for this. I consider myself relatively savvy with the Big Three testing sites as well as triangulation; however, I did not realize 23andMe permitted triangulation.

    I wonder if the reason why so many of us don’t know about this feature is due to the way it is labeled. When I read a “Yes” in the “Shared DNA” column, it sounds like they mean the person is open-sharing, or just sharing with me, in general. Perhaps it could be re-labled “Same segment?” or something more intuitive.

    Liked by 1 person

    • Micaela
      I noticed what was happening when I went down each ICW list – the Yes’s were always in a TG. Then I noticed the little info icon beside Shared DNA – click on that to get their “overlap” explanation, without mentioning Triangulation.

      Liked by 1 person

  9. Pingback: Triangulation at 23andMe | Maggie's Genealogy Blog

  10. Jim, In the 23andMe ICW list, ALL the people on the list are shared relatives, even if we don’t share exactly the same segments, right? Seems to me that’s a helpful clue, even if it’s not triangulation of a segment. For instance, I see a known 2nd cousin on my mother’s father’s side in one list, but he is a NO. I would take that as indication to look for a connection in that line, especially if a number of names that I have in a TG show up in that same ICW list. Am I misunderstanding something?

    Like

    • mizmdk,
      Yes. All you’ll ever see at 23andMe and FTDNA are your Matches. An ICW list is a subset of your Matches. I’m not sure how to use an ICW list effectively – by itself, it doesn’t tell you much. A lot of folks have gotten really confused by it and thought is was a TG list – NOT. A TG is ICW AND on the same segment. So a subset of the ICW list would be those that also have overlapping segments – that’s what the 23andMe “Yes” means.
      Sometimes a close cousin in an ICW list may have the same ancestry with the base Match that you have, but your close cousin has lots of other ancestry than the Common Ancestor with you, and could well be matching the base Match some other way. IMO, an intermediate goal should be to group all Match-segments into TGs. At least then you know they share a Common Ancestor. And some different TGs can have the same Common Ancestor, just on a different segment. Close cousins often share more that one IBD segment with you, and usually (but not always) you will have the same Common Ancestor, or more accurately the same MRCA, with that cousin.
      I don’t think you are missing anything – in addition to Triangulation, there are clustering, FAN methods, mirror Trees, researching a Match’s Tree, etc. – lot’s of ways to tackle the genealogy side of this – finding the Common Ancestor.

      Liked by 1 person

      • “IMO, an intermediate goal should be to group all Match-segments into TGs. At least then you know they share a Common Ancestor. ”

        I’ve been led to believe that triangulation does not prove a common ancestor (IBD) and that there aren’t studies that show that triangulation alone increases the likelihood that the segment is IBD. For this, don’t we need to go by a) segment size, and b) phasing?

        Like

      • Micaela, Nothing “proves” a Common Ancestor. Triangulation is a tool that will add confidence to a Common Ancestor linked to a DNA segment. Just finding a census record or a marriage licence or deed or Will, or all of these, does not “prove” a CA. And you are right again, in that I don’t know of any studies about Triangulation alone. I do have my own experience with over 8,000 individual Matches. With Triangulation, I’ve culled out over 1,000 “shared segments” reported by the companies as not Triangulating where they should if they were IBD. For me that means a much higher likelihood of IBD than what we get from the companies. Although I have not published a paper, I also have not found any of my Triangulated segment to be false. The TG does not prove a CA, the genealogist has to determine the correct CA for each TG. Segment size often offers clues, but DNA is very random with a wide range of possibilites. Except for very close relatives, we cannot prove any relationship with segment size. Phasing is great if you have one or two living parents. Many genealogists do not. And phasing does not prove a Common Ancestor.

        Like

  11. I’ve run into this issue:
    Match 1 – shares a segment on CHR 1, 4, and 9 with me
    Match 2 – shares a segment on CHR 1, 3, and 9 with me.

    Using match one’s ICW list that ‘yes’ could be a shared match with 1 and 2 on CHR 9 , and thus point to a different common ancestor than a match on CHR 1 would, which would throw off triangulation results.

    I’ve been making sure that the ICW lists I use to put matches in only have 1 segment shared, so those ‘shared dna: yesses are guaranteed to be correct.

    Am I overthinking this?

    Like

  12. I was following the instructions on the blog but I am finding that many (most) of the ‘Yes’s match the person I am forming the TG for match them on a different chromosome (not the one where I am making the TG) and so are not in the TG. By the time I have clicked on them to find this out and then go back to check the next ICW have lost the page I was working through (eg 7 of 12) and then have to click several times to get back to that page. It is all rather tedious. Perhaps I could get round this by having more pages open at once. I was under the impression from the blog the Yeses would match the TG. So I am not sure of the difference between the YES and the NO.

    Like

    • All the Yes’s should be in the same TG with you and the base Match. It is best to look at lists with Matches who share only one segment with you, and create the one TG with that Match. It you use someone with multiple shared segments as the base, you will be working on multiple TGs at one time – I should make that more clear. An option for you if you want to use a multiple-segment Match as the base, is to just write all the Yes’s on a piece of paper and then put them into groups by segments later. Or, if you start with a spreadsheet of all your Matches, sorted by Chr and Start, then just skip over the Yes’s that aren’t in the TG you are working on, and come back to them when you are working on the next TG.

      Like

      • I started with the first match on chr 1, who happened to be a known 2nd cousin. So I went through all his ‘yes’s to get the hang of it and marked them on the spreadsheet (my aggregate download rather than my main one) on their respective chromosomes, so I would not have to go through them again. I did all the TGs on chr1. I realised it was going to take days to do all the chromosomes. So I then started on chr 11, which was shorter, and also had known cousins as these TGs with known cousins are useful for mapping. I realise it is not much use using my 1st cousins for this as I already have used them for Maternal/paternal sides and would not get more information from them.
        With the 2C2R on Chr 11, with whom I have several shared segments, I again noted each match’s Yes match and I noted a different TG code for each chromosome, which I can tidy up later when I get to that chromosome.
        Which brings me to my question. How do you deal with the different ancestral levels? Eg my 2nd cousin gave lots of large TGS -01A etc. A 3rd cousin, related to that 2nd cousin’s parents, will have matches who are in 01A, but they are a subset, and they will also have matches not in 01A. And a 3rd cousin on that 2nd cousins’ other line will have a different subset. How do you code these?
        Thank you for all your blog comments and these articles.
        Sarah Jordan

        Like

  13. Sarah; for a Match with multiple segments: first they are in my spreadsheet once for each segment with a notation like 1/3, 2/3, 3/3 – with 1/3 being the largest segment (it’s a tell-tale to remind me of the other segments. Then I reference back to the other segments, e.g. on segment 2/3 I note: [1/3= 04G25] – this means the 1/3 segment is on Chr 04; starting in area G; on my father’s (2) mother’s (5) side. As you know, most multiple segments will be from the same CA, but not always. So sometimes these notes denote a conflict. The 2/3 segment above may be in TG 13D25 – no problem; but it might be in 12F24 – indicating father’s father’s side – heads up! [an appropriate term for the time of night when this often occurs] – this means either 2 different CAs or one of the TGs has an incorrect label (I’m not hesitant to change a TG label – they are tools that work for me) so I color the background in these notes red to highlight that the TG ID or CA has to change. In some cases the two TGs are on opposite sides (one segment matches a maternal uncle; the other my dad). My TG letters, generally, refer to where the TG starts: A for 0-10Mbp; B for 10-20, etc. – there are just enough letters for Chr 01, but I’ll use Ga and Gb, if that’s where a TG subdivides. This isn’t are hard rule (TG IDs work for me, not the other way). I just make sure they stay in order, so that a spreadsheet sort, by side & TG ID will keep them in order within a side.

    Like

    • Thanks so much, Jim. My natural inclination (I was a scientist!) has been to be as organised like this – from when I first read how you numbered your TGs, but I found it a struggle to get my ageing brain around it and I got tempted to simplify. Especially as I am in the UK and have great big gaps due to having fewer matches. I will certainly use your way of noting matches on multiple segments as it is neater and quicker than my current method. Also, I will use the lower case letters for when the TG subdivides. If only there were more hours in the day! Having only just been transitioned to the New Experience the sudden influx of many more matches has overloaded my routine of updating Family Finder matches and contacting Ancestry matches. Updating spreadsheets for tested cousins has fallen by the wayside, though I do try and keep, my husband’s, FF and Ancestry data going. At least I have ADSA back which I find a big help for FF.

      Like

    • Jim, let’s say you get a new match from 23andMe, and their segment falls in between or overlaps two TG’s you have already named/numbered. How do you code/number this? I feel like there is some basic concept I am not getting here.

      Like

      • mizmdk, Well first of all this rarely happens. If a new shared segment truly spans two existing TGs (that is the segments in both TGs match the larger segment), then that means the new segment is from a somewhat closer cousin (close cousins often have shared segments that span TGs). If a new shared segment only falls into the next TG a little, ignore that fact – chalk it up to fuzzy data – you’ll find the new segment doesn’t Triangulate (over 7cM) with both the existing TGs. Also make sure the two TGs that you are “overlapping” are on the same side as the “new match” – in other words a TG on your maternal side could easily overlap with any combination of TGs on your paternal side – they are completely independent – on different chromosomes…
        Jim

        Like

  14. What am I missing? When I do this, of the 1203 matches that end up on the downloaded spreadsheet, only 2, yes 2, have any segment data.. What does this mean. These are the 2 people that I have a “connection with.” Every other person, the segment data is empty.

    Like

    • Becky – I cannot be sure, but I’d guess that you have not selected “Open Sharing” at 23andMe. You need to share your DNA/segment data with everyone at 23andMe, in order to have shared segments show up in the segment download. Jim

      Like

  15. I think I have this, but need a little clarification. I’m starting at the top of my sorted list with Match A. Match X is a Yes, but Match B is No. There are several more Yes Matches, all of them (along with A & X) in my first TG. Now I go to Match B, and I see Match X on the ICW list with a Yes. Does Match B belong to my initial TG, or does it begin my second one?

    That is: A/X=Yes (A & X both in my TG 01A); A/B=No (B not in TG 01A — at least, not yet);
    but B/X=Yes. Does this make B part of TG 01A, or is X in both 01A and 01B? I’m guessing that it’s the former (that is, the relation is transitive), as you make no mention of a match being in multiple TGs.

    Like

    • Followup question: I just did a re-sort, adding a third sort key of distance (descending), which put my sister at the top of the list, and all but one of our common matches are Yes. It seems that I should probably omit my sister from the list… or, am I missing something here?

      Like

      • Alfredo, I’m not sure what you did, but with an ICW list with one of your Matches (the base Match), the Yes Matches should all fall into (triangulate with) the segment of your base Match, including your sister (NB: if you share more than one segment with the base Match, the Yes’s may be spread over all the segments you share). The Yes or No designation should not change – the order doesn’t make any difference. Jim

        Like

    • Alfredo, In this case you have to look at how much overlap there is. If there is little overlap between A and B, that would explain why there is no match. And B could be in the TG, just at the other end from A. If B overlaps A so that you’d expect a match if they both got that segment from the same Ancestor – then if no match, I’d start a new TG with B. As you add Matches to TGs these apparent anomalies sort out. In general a segment form X cannot be in two TGs – except when X is a close cousin, with a large segment, can span more than one TG. In general, most of your shared segments will “fit” into one TG. If it seems awkward, skip that segment and come back to it later. Jim

      Like

      • Thanks for the reply. In my initial question, the ‘X’ was my sister, so she ends up being part of just about every match. What this seems to be saying to me is that I probably shouldn’t include my sister for this type of analysis, as she is too close (kind of like a geographical triangulation with two of the points just inches apart when trying to measure miles).

        I do have a 2nd cousin once removed who may be a better candidate for triangulation (if that’s what I’m doing in this case): I have two other matches, Orange and Yellow, who both have common shared segments with me and 2C1R (Yellow actually has two, on separate chromosomes), but no common shared matches between Orange, Yellow and me. Orange’s surname list includes the maiden name of my great-grandmother (the maternal half of the CA pair between myself and the 2C1R, and it’s not a very common name, so it’s highly unlikely to be coincidental), so I assume that Yellow is more likely to be related via my great-grandfather (JP) instead. Now, Yellow doesn’t recognize JP’s surname in her tree, so my guess is that the relation would be through JP’s mother or one of his grandmothers (my shared DNA with Yellow is 135 cM, with a largest segment of 62, so she’s probably not much more distant than 3rd cousin). Does this sound like the kind of thing you’re using triangulation for, or would this be something else?

        In any case, I’m looking forward to going through my entire list; the process itself is fascinating, and if it helps uncover more relatives, even better!

        Like

  16. Jim,
    Your post is very interesting. As others have expressed, I had not properly understood the meaning of “Shared DNA”. So this article was quite a revelation to me.
    I am trying to follow your procedure and I have a question. I start on Chromosome 1 and mark the first match as 01A and then search down the Relatives in Common list on 23andme.com and wherever I see a “Yes”, I mark the corresponding name in my downloaded spreadsheet (greg_simkins_relatives_download (54)). I match this relative on multiple chromosomes and don’t know how to discern which chromosome the various matches share with me, so I am marking match 01A on many chromosomes other than 01. I guess I will sort them out later using the Chromosome browser in Genome Mate Pro. Or do you suggest a simpler procedure?
    Thanks,
    Greg

    Like

    • Greg – In creating Triangulated Groups, each one is on one segment on one chromosome. Think of a long brick wall only two bricks high – one row is your maternal chromosome and the other row is your paternal chromosome, and each brick represents a separate TG from a specific Ancestor, coming down one line of descent to you. So TGs do not/cannot jump chromosomes. If you have a close relative who shares long or multiple segments with you, you have to adjust the procedure a little bit. The 23andMe “Yes”s are usually sprinkled over all the shared segments from a close Match, so if you are working on Chr 01, you need to make sure the “Yes” Matches have an overlapping segment on Chr 01 with you. Usually, if you use a spreadsheet, or GMP, with a sort on Chr and Start – this is not a problem at all. You can see the “Yes” Matches nearby in the sorted data. If the “Yes” is not nearby, it’s probably on a different segment. I have that virtually ALL of the “Yes” Matches do fall into TGs. Large segments may span two, or more, TGs. Use your judgement and visualize those bricks. Jim

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s