Some DMT Dialog, with Questions and Answers - Sat, 12 Nov 2022

Michael Kaplan, a user of my Double Match Triangulator program sent me information about how he was using DMT and also asked a number of questions. I thought this conversation would be very good for other DMT users to read. Michael gave me permission to post the relevant parts of his email to me and to give his name. Thank you Michael.

Here is what Michael had to say and ask (shown in “quotes”), along with my responses in green.

The Dialog, Questions and Answers

“I have begun using DMT (on GEDmatch) in earnest.  I now have a total of 161 kits in my B folder, including your kit.  Note that I do have a number of unrelated people in the database, including my wife, our daughter from my wife’s first marriage, and a couple of adoptees that I have helped. I’ll also add unrelated matches of my own matches, in order to determine of which side of my match’s family our relationship might be.”

“I began the process by generating my own segment match file, then doing the same for each of my known cousins.  When doing the segment search, I left the default minimum segment size as 7 cM.  That’s recommended, isn’t it?  I don’t want to bother with false matches.  In addition, reducing that threshold would actually reduce the number of discrete matches captured, since the limit of 10,000 segments would be reached much sooner.” 

Yes. The default of 7 cM is recommended. Even triangulations can be false (matches by chance) if they are under 7 cM.

“I manage kits for quite a few of my cousins.  Others have done the upload to GEDmatch on their own.  Then assigned MRCAs, as appropriate.  I do manage a kit for one close match whose relationship is yet to be determined.  Renee Watson has ended up in the "FFF" cluster but haven’t confirmed the exact connection.“

“Is there a recommended next step?  Unfortunately, I don’t think I documented the exact path that I have taken. I’ll try to reconstruct my thought process.

1. Find the closest match (person B) which has not yet been added to the B folder.
2. Generate segment match file.
3. Make initial DMT run to generate People file for person B.
4. Re-run DMT on my People file (person A), looking for triangulations with known cousins (and other matches already processed)
5. Run "Match both kits or 1 of 2 kits" on A and B, with shared matches first ordered according to shared DNA with A, attempting to find additional triangulations. 
6. I’m running DMT with recommended settings of 7 cM for Min Triang and 15 cM for Single Triang.  Default settings for "Match both kits or 1 of 2 kits" are 10 cM for threshold of largest segment to qualify as a match, and 10 cM for threshold of total matching segments to qualify as a match.  Should I reduce both of those settings to 7 cM to match DMT?
7. Then select the closest shared matches (ideal number TBD) for "Multi-Kit Analysis", running "Triangulation" on that subset.  Default setting for triangulation is 7 cM.  Should I leave it at that? 
8. For each shared match C which is reported as triangulating with A and B, generate a segment match file and add to the B folder, if one has not yet been created. 
9. If the segment match file had already been created for C, confirm that B is shown as a match in C’s People file.  If not found, generate one-to-one files for match between B and C.
10. Then re-run "Match both kits or 1 of 2 kits", this time ordering shared matches according to shared DNA with B.
11. Repeat steps 7-9.
12. Now, look at tab on "Match both kits or 1 of 2 kits" for kits matching only B.
13. Generate segment match files for the closest matches.
14. Re-run DMT for person A (me) and person B.
15. If tree for person B is available on GEDmatch, try to assign MRCAs on B’s People file
16. Otherwise, examine source for GEDmatch kit (e.g., Ancestry, MyHeritage, FTDNA) and check if tree available on that platform.
17. Return to step 1 and iterate through remaining unprocessed matches, periodically re-running DMT, as appropriate.”

Step 6:  No need to change default GEDmatch settings. The 7 cM and 15 cM settings in DMT are for individual segments. The 10 cM settings at GEDmatch are for longest segment and total of segments to determine which people to include as matches and does not filter individual matches. With endogamy, you might if anything want to increase the longest segment, rather than decrease it.

Step 7:  Default setting for triangulation is fine at 7 cM.

Step 12:  I’m not sure why you want kits matching only B. The point is to find the connections with Person A, so how will anyone not related to Person A help?

Steps 13 to 16:  I don’t see how using DMT for the B people is useful unless you want to determine Person B’s connections. Looking up Person B’s tree is definitely useful to see if you can find a connection with Person A’s tree, but there really is no need to find person B’s MRCAs and run them as person A in DMT if they are not your subject of interest.

“What’s the optimal procedure to determine the next kit to add to the B folder, given only the available kits in GEDmatch?”

I don’t know if there’s an optimal way. Using the closest match as you are doing is okay.

“How do we deal with a match with status "In Common With B" in assigned cluster "M"?  Do I need to find shared matches which will triangulate?”

“In Common With B” indicates that A has segment matches with C and B has segment matches with C, but none of the matches triangulate for A, B and C together. To find triangulations, you’ll need segment match files of other B people who have matches with C on other segments.  Matches will triangulate for everyone who is a true match of A on that segment on one of the parent sides.

“For status "Only AC matches", I need to also find shared matches which will triangulate.  Is that true?”

In this case, person C does not match any of your B people. The segments Person B matches with you should not triangulate with anyone else on both your father’s and mother’s side. If they do, then either the segment is false or your father or mother’s triangulation group must be false. As you fill in more and more of your segments with triangulation groups, you should reduce the number of “Only AC matches”.

“Matches with status "Has triangulation" in cluster "U" appear to match multiple other kits at the same location.  However, those segments have not yet been associated with a common ancestor.  Is that correct?  I assume that process involves more in-depth examination of family trees.” 

Yes. If none of a person’s segment matches have been assigned an ancestral path, then they will be put in Cluster “U”.

DMT can only assign ancestral paths to segments where the MRCA of Person B or C is known. So the goal is to include as many MRCAs as possible in Person A’s file, whose relationship with Person A is known and is only a single relationship (not related 2 or more ways). Second cousins and further are best since they will give you at least the grandparent level FF, FM, MF and MM.  It might sound like you’d want to go as deep as possible, e.g. 6^th cousins, but the worry the farther out you go is that they might be related a different way that you don’t realize.

“While I am aware that I have a large family and many known cousins have already tested (and uploaded their data to GEDmatch), those kits do not uniformly cover all of my ancestral lines.  There are still quite a few gaps in my chromosome map.  Importing the chromosome map data DNApainter.com, on my paternal side, I’ve got 46% of 396 segments painted, including the following:

On my mother’s side, though, it’s only 13% of 131 segments painted:

Total is 29% of 527 segments painted.”

“Do you have any recommendations as to how I can increase that test coverage?  My father’s side is larger.  I’ve just sent my sister an Ancestry DNA kit.  Hope she might be a closer match to some of my more distant matches.  I’ll transfer her data to GEDmatch when it’s ready.  Another paternal 1st cousin has already tested on 23andMe and Ancestry, I’ve asked him to upload to GEDmatch.  One of my maternal 1st cousins is already on GEDmatch.  I’ve asked the rest of my 1st cousins who have not yet tested anywhere if they’d be willing to contribute to my ongoing research.  Unfortunately, one of those was adopted, so not a blood relative.  Another has had a bone marrow transplant and is not a good candidate for a DNA test.”

Once you’ve got all your relatives tested that you can get tested and you’re maxed out on how the % you have painted, then it might be worthwhile to study your DMT Combined Run for Person A (yourself) to see what else it tells you. Start to identify triangulation groups the way Jim Bartlett does on his segmentology.org website. 

Another thing you can try, is to go into your People file and look for people who match on 4 or more segments and look for people whose segments are all in the cluster they are put into or else are “U”. That may indicate (no guarantee, especially with endogamy) that they match Person A along that cluster’s ancestral line.  Then you can try adding MRCAs for them, but don’t include the “R” at the end, since you don’t know their exact MRCA with person A. e.g.  If they have 4 segment matches on FF, FF, FFM and FF, then you try assigning them an MRCA of FFM (not FFMR). DMT will use that MRCA as part of its attempt to initially assign each segment, but it will be overruled by any triangulations with ancestral paths from MRCAs ending in “R”. This may help fill in some unknown areas where there are no known relatives with segments.

“Is it feasible to work the other way around, say, using the DNApainter chromosome map to select an as yet unidentified segment on one side or the other?  Then enter that segment into the GEDmatch segment search tool, finding matches who have a shared match with me on that segment.  I could then examine trees for those matches, if any, to potentially find a common ancestor.  If no tree is available (or insufficient depth is provided), I could reach out to the match for additional detail.” 

You can do that to find people at GEDmatch who match you on a specific segment. The only problem is, you won’t know what ancestral path they match you through. Searching that person’s ancestral tree will only help you if they connect to Person A’s tree. Everything in DMT including the MRCAs are done from Person A’s point of view/

“I have begun building floating trees for many of my closer unconfirmed matches on Ancestry, where they have trees.  Also using public records searches to identify other matches who have relatively unique names.  I’m reaching out to my closest matches, introducing myself, listing my known immediate ancestors, describing our shared matches, offering to share trees, and asking of they’d be willing to upload to GEDmatch.  I’ll do the same on 23andMe, MyHeritage, FTDNA, and GEDmatch.  I’ve given up on LivingDNA.  Response is likely to be low, but I’m hoping to find connections to the poorly covered branches of my family.  I’ve already gotten a favorable response from one close match and have begun to analyze our connections.

As you are progressing through an analysis, at what point do you give up and determine that a match is likely due to endogamy alone and probably untraceable?” 

In my case, any match that goes further back than my furthest researched ancestors is probably untraceable, simply because there are no more records from the Russian Empire and Romania that will connect them.

With regards to endogamy, it is almost the same as pileups. Endogamy segments are from long ago that have been passed through many people on various lines. They do indeed triangulate because they are actual segments and each segment is passed down to you through one specific ancestral path. But that path is always too deep to trace and connect to.  They are no different than any match that goes further back than my further researched ancestors.

“What are your minimum criteria for considering a match in the first place.  I generally focus on matches with more than 90-100 cM shared DNA and largest segment of at least 15-20 cM.  However, I’ve had cases where ThruLines (on Ancestry) has found a legitimate match of as little as 8-9 cM, when there are well-documented trees on both sides.  So, I’ve tried to provide as much depth on my tree as possible.”

For the B people, I’ll start taking the people I know the MRCA for. For the rest, I’ll then want to take the highest matching ones in each cluster, and try to get 5 or 6 in each cluster so that I have maybe 50 Person B files. I’m not as concerned about the cM as much as trying to get coverage of the closest matches in each cluster.

“Does DMT make any adjustment for known pile-up regions or any other corrections for endogamous populations?”

No. A pile-up or endogamous region are true segments that are passed down an ancestral path. The only difference is they are too far back to find the connection. You can identify them easily in DMT when you see dozens or even hundreds of triangulations all lined up over a segment. These could overlap with perfectly valid segments of closer relatives who can be identified, so don’t ignore those regions. Just ignore the people you don’t recognize in those regions.

“Any further strategies/hints/suggestions for making optimal use of your wonderful tool, DMT?  I’m very happy with it, so far, and hope to learn quite a bit going forward.”

Check out the last page of the Help File that comes with DMT, titled:  “Ideas, Tips and Tricks”.