Triangulated DNA matches and Pile-up Areas #dna


Lee Jaffe
 

I would appreciate help to determine whether a cluster of triangulated DNA matches is significant or whether it is negated because it is in a pile-up region.  
In the example from MyHeritage, I've identified four people who match me on the same segment – Chromosome #2   88,474,256 - 105,231,018
All of the individuals have fairly strong matches to me, enough that I already considered them higher priority on their own.  The possibility of identifying four matches with a suggestion that we  share a common ancestor might be a useful clue in solving some of my brick walls.



However, I recently attended a DNA workshop where the instructor mentioned that matches in pile-up areas were not useful.  According to the chart I copied, there is a common pile-up area on Chr #2 starting 85,304.243 and ending 99,558,013  As you can see there is some overlap with the triangulated segment I am investigating.  BUT, the triangulated segment isn't completely within the pile-up area where the examples shown during the workshop were entirely within the pile-up area.  Therefore, I wonder ...

Is the triangulated segment significant or not? 

I've found a number of good guides about pile-up areas and also those discussing triangulation but I haven't managed to find anything which relates one to the other.  I'm hoping members of the list can answer this or point me to a source which provides the answer.  I quit Facebook a year ago, so I don't have access to the relevant groups there.  I've tried to join JewishGen's DNA Testing mailing list but haven't yet managed to navigate that resource.  

Thanks in advance for your help,

Lee David Jaffe
===============
Surnames / Towns:  Jaffe / Suchowola, Poland ; Stein (Sztejnsapir) / Bialystok and Rajgrod ; Roterozen / Rajgrod ; Joroff (Jaroff, Zarov) / Chernigov, Ukraine ; Schwartz (Schwarzstein) / Ternivka, Ukraine ;  Weinblatt / Brooklyn, Perth Amboy, NJ ; Koshkin / Snovsk, Ukraine ; Rappoport / ? ; Braun / Wizajny, Suwalki,  Ludwinowski / Wizajny, Suwalki

 


Adam Cherson
 

Lee,

I'm not a 'degreed' geneticist but have been working as a private researcher for many years. What I am about to say is only my perspective based on experience. In addition to the possibility of pile-up zones, there is another aspect and that is the length of and number of triangulating segments. In your example I see a 6.6 cM segment and do not know whether this is the only one for the group.

My view is definitely in the 'more is better' camp. In my work I tend to disregard anything under 7 cM and especially if it is the only triangulating segment for the group.This is all the more so when one considers that many of the chip-reading programs use something called imputation, which is in effect a way of filling in gaps between segments to make them appear longer.

As with all genetic analysis I feel much more secure when triangulation results are supported by other evidence, both genetic and non. For example I tend to be accept smaller triangulation segments when the overall amount of matching is consistent with the degrees of relationship I am trying to prove for the various group members.

Genetic analysis is often somewhat impressionistic and I don't think anyone can say for sure that this or that segment length is or is not significant. I look at your example this way: if you remove the pile-up zone from consideration, you have a triangulation of about 2.4 cM, which is below what most people, including myself consider significant. Therefore, if this is your only evidence of common ancestry for the group I wouldn't consider this result as proof. If it is only one of several other types of evidence then it could lend a tiny bit of support to the hypothesis.

There is a blog and discussion dedicated to segment analysis which may discusses such matters as segment length and pile-ups in greater detail and may provide you with valuable information: https://segmentology.org/ You may want to re-post your question there for additional opinion and discussion.

--
Adam Cherson


Ellen
 

Lee,

I believe that I attended the same DNA workshop that you did.  I haven't had a chance to try it yet, but the instructor suggested using DNA Painter.  My understanding was that DNA Painter will automatically identify the pile-up regions so you can exclude them.    

I agree with Adam.  Even though MyHeritage is showing triangulation, if the shared segment is less than 7 cM, I would disregard it.  

Ellen 
--
Ellen Morosoff Pemrick 
Saratoga County, NY

Researching WEISSMAN/VAYSMAN (Ostropol, Ukraine); MOROZ and ESTRIN/ESTERKIN (Shklov & Bykhov, Belarus); LESSER/LESZEROVITZ, MAIMAN, and BARNETT/BEINHART/BERNHART (Lithuania/Latvia); and ROSENSWEIG/ROSENZWEIG, KIRSCHEN, and SCHWARTZ (Botosani, Romania)


Phil Karlin
 

Given the length shorter than 7 cM, I agree it's probably not worth pursuing. I don't disagree with anything Adam or Ellen already wrote. 

That said, I don't think I've ever seen in my results a 5-way triangulation. If you drop #2 (orange) does the triangulation with the others lengthen? How about if you drop #4 (green) as well? 
I think a long shared segment in a pile-up region is still a long shared segment. It just may be from a common ancestor further back than the same size segment in a non-pile-up region, and so less predictive. 

Final thought: whether researching something is "a waste of time" depends on how much time you have and what else could you be spending it on. If the prize is big enough, the investment may be worth it, even if it's a long shot.

--
Phil Karlin
Hartford, CT USA


Lee Jaffe
 

I very much appreciate the suggestions and comments I've received here and privately.  I think I understand the approaches suggested, have looked at additional guides and sources, and have pivoted, I hope, to a better strategy. 

Before getting into the new approach, I wanted to respond to one note which posed whether this was the best use of my time and energy.  Honestly, I don't know and I hope that exchanges here can help me figure this out before I go to far.  My goal is to extend my family research beyond what I've been able to achieve so far through conventional or traditional methods.  I have a fairly well-developed tree,  which goes back in a couple of cases 9 generations.  But there are some mysteries therein, such as the parentage of my 3x great-grandmother or what happened to my great-grandparents' newly discovered siblings and their descendants. I am hoping that triangulation will help me distinguish DNA matches into groupings according to possible common ancestors with the goal of placing them in my family tree when I follow up with traditional methods.

Therefore, following new sources, I have downloaded my match data from FTDNA, 23andme, and MyHeritage.  I've started to filter and sort the entries by Chromosome number and starting position in order to identify potential triangulated matches.  But the number of entries (~150K lines) is cumbersome and daunting.  Just starting with the beginning of the MyHeritage data, I have more than 100 matches starting at the same position on Chromosome 1.  Given that MH allows you to compare only 7 matches at a time in its Chromosome Browser, I've been trying to find ways to prioritize which would be most likely and most useful.  It's not straightforward.  For instance, I've discovered that a batch of segments with the same beginning and ending location do not triangulate, at least not according to MyHeritage's Chromosome Browser even though it says it allows for segments as small as 2cM.
Chr# Start Location End Location Start RSID End RSID cMs SNPs
1 752,566 4,007,008 rs3094315 rs7519349 7.2 1920
1 752,566 4,007,008 rs3094315 rs7519349 7.2 1920
1 752,566 4,007,008 rs3094315 rs7519349 7.2 1920
1 752,566 4,007,008 rs3094315 rs7519349 7.2 1920
1 752,566 4,007,008 rs3094315 rs7519349 7.2 1920
1 752,566 4,007,008 rs3094315 rs7519349 7.2 1920
Plugging combinations of the above matches from MH into its Chromosome Browser gets a "no triangulated segments" result.  Note: all are above the 7.0 cM threshold recommended in the comments I received. 


Then I looked for other groupings featuring larger segments and had better success with those with segments larger than 10cM.  For instance, in the following batch, all but two of the matches triangulate with each other.  A and F are the exceptions, even though the numbers for A are identical to B, as is F to E.  

A 1 10,825,577 17,954,411 rs11121615 rs12060961 14.6 4096
B 1 10,825,577 17,954,411 rs11121615 rs12060961 14.6 4096
C 1 10,825,577 18,279,185 rs11121615 rs709362 15.3 4352
D 1 10,825,577 18,528,026 rs11121615 rs9725311 16.1 4608
E 1 10,825,577 18,528,026 rs11121615 rs9725311 16.1 4608
F 1 10,825,577 18,528,026 rs11121615 rs9725311 16.1 4608
G 1 10,825,577 19,022,911 rs11121615 rs12563055 17.7 4992
H 1 10,825,577 19,136,610 rs11121615 rs11261075 18.4 5120

I feel like I need to understand why I'm seeing results like this before going much further down this path.  As always, I appreciate your help.

Thank you,

Lee David Jaffe
===============
Surnames / Towns:  Jaffe / Suchowola, Poland ; Stein (Sztejnsapir) / Bialystok and Rajgrod ; Roterozen / Rajgrod ; Joroff (Jaroff, Zarov) / Chernigov, Ukraine ; Schwartz (Schwarzstein) / Ternivka, Ukraine ;  Weinblatt / Brooklyn, Perth Amboy, NJ ; Koshkin / Snovsk, Ukraine ; Rappoport / ? ; Braun / Wizajny, Suwalki,  Ludwinowski / Wizajny, Suwalki

 


Adam Cherson
 

"The same question seems to come up over and over again among those new to autosomal DNA testing. If I match A and B on the same segment why is that not enough to prove they match each other and we have a common ancestor?

The reason the ancestor is not proven is that you have two strands of DNA on each chromosome (remember there are 23 pairs of chromosomes) and the testing mechanism cannot differentiate between the two of them. So A could match the piece from your mother and B could match the piece from your father or one of them could even be a false match to a mix of alleles from both parents (see my post on IBC for more on that concept)" from http://blog.kittycooper.com/2015/02/triangulation-proving-a-common-ancestor/ (this is an old post so ignore the techniques shown on the rest of the post)

In the first example, every member of the group matches you but not each other. In the second example the group matches you and they all match each other.

To do triangulation manually you need to see not only who matches you, but then also compare them to each other and see if they match at the same locations. The 3-D Chromosome Browser on Gedmatch gives you a table showing the internal matching of every kit you put into the group (note that the table duplicates each match by reversing the kit order). If you sort this output by chromosome number and then by location you can then see quickly whether there are any identical positions of internal matching between more than any one pair in the group. If you have Tier1 tools there is an automated triangulation app available, which is even faster, and safer. The MyHeritage triangulator also works well with up to seven. I'm not sure what the number limit is on Gedmatch. I do not believe FTDNA, Ancestry, or 23 have triangulation, although I am not up on the latest platform upgrades.
--
Adam Cherson


Eva Lawrence
 

 

Like Lee, I've got really interested in the theory of triangulation. Ive been trying to think if it in the abstract as a mathematical problem.

.In the abstract. it just depends on two lines of ancestors meeting in a single point, another ancestor  - but there's no such thing as single ancestor. It takes two to make a baby.  Each point on your tree is a node,  A person's DNA is just the track of one line of their enormous tree,  going back to 'Adam' or a Neanderthal man but also to  Eve or Neanderthal woman..

Triangulation can only find a node from their ancestral line which also lies on someone else's ancestral line if the line is reflected in their DNA -or in their paper records..

Geometrically a line is defined by two points. Lee put the (very partial) back history of 11 people, A to G who share one segment of DNA into a black box - the algorithm used - and it searched the rest of their DNA to find a second piece (a series of numbers representing  chromosomes)  that existed on one of the other sets of DNA.. . ie a second point where the lines cross.  But it's no surprise if the algorithm doesn't find one, because its existence is almost independent of the first piece - like the second throw of a dice theoretically doesn't depend on the previous throw

The further back you go, the more widely the net of ancestors has spread,  but only a small proportion of  lines is recorded as leading  back to each of the initial list of matches (who were people on a database, but nodes in the abstract)  and provides the information on which the triangulation algorithm reports.

My second  thought was that it was not just a single line, but as Adam Cherson pointed out, at some nodes two lines back are possible because both parents have inherited that piece of DNA. But unless they were identical twins (!)  there must be some difference in their make-up that takes the DNA line back unequivocally to another node. 

It is only because communities were limited in size  that triangulation has any success.

Eva Lawrence

St Albans, UK

 
 

--
Eva Lawrence
St Albans, UK.


Lee Jaffe
 

As a belated follow-up to this discussion, I want to say that I've decided that using triangulation effectively is outside of my wheelhouse.  I've looked at the guides, taken many seminars, followed the suggested strategies, prioritized my matches, created tables and chromosome maps, and stared at tens of thousands of data points and ... nothing.  I won't go into all of the hurdles and traps I've encountered trying to establish whether or not parties share triangulated segments: generally, just because sets of segments look triangulated doesn't mean they are, and vice versa.  And, to top it off, even when you have identified sets which are triangulated, it isn't clear that they are significant.  

I'm not a naive user.  I've added hundreds of members to my family tree, prompted by DNA matches.  Yet never once has triangulation played a significant role nor has DNA made a conclusive case. Instead, I've come to think of like the game "Marco/Polo".  When you post your DNA test results, you are essentially calling out "Marco" and listening for someone else's test to respond "Polo."  You have a vague idea of distance and direction, but success ultimately does not depend on further DNA analysis, at least not alone.  Over and over, it's been a case of a match – often those which fall below the thresholds experts recommend – highlighting a person whose family surnames or town name or some other clue suggest further examination.  Then it's a matter of cooperation, sharing trees or offering to ask an aunt who might remember ... and one day they send you a photo of a mutual great-great-grandparent.  (It's happened, more than one, really.)  But without the cooperation, all the DNA evidence has gotten me nowhere.  


Lee Jaffe
===============
Surnames / Towns:  Jaffe / Suchowola, Poland ; Stein (Sztejnsapir) / Bialystok and Rajgrod ; Roterozen / Rajgrod ; Joroff (Jaroff, Zarov) / Chernigov, Ukraine ; Schwartz (Schwarzstein) / Ternivka, Ukraine ;  Weinblatt / Brooklyn, Perth Amboy, NJ ; Koshkin / Snovsk, Ukraine ; Rappoport / ? ; Braun / Wizajny, Suwalki,  Ludwinowski / Wizajny, Suwalki