How can I interpret multiple small segment matches?

+1 vote

My sister shared DNA with a person, X, on Ancestry.  When I looked at that person’s tree I recognized a name and so went to GEDmatch since X’s profile showed her kit number.  I was able to contact X’s son who with another member of the family helped me understand the relationships between X and 6 members of her family who matched my sister and me. The family members were related to X as son, daughter, niece, 1st cousin, half 2nd cousin, and 1st cousin 2x.  I did a one-to-one comparison between all combinations of these 7 family members, my sister, and myself 9 ( 36 comparisons in all) and found that except for one combination, all of them shared overlapping segments on Chromosome #9. The issue is the segments were very small - only between 4.0 and 5.0 cM.

How likely is it that I’m related to this family? 

in Genealogy Help by Anonymous Loughlin G2G Crew (740 points)
Triangulation has not come up as part of this discussion but it seems relevant and may be a useful tag to add. Barry has commented on my similar question (search g2g '41') and another person, outside WikiTree, has suggested 'sticky' (persistent) segments may extend matches much further back than atDNA normally reaches. He suggests taking care when there are more than say 20 small segment matches. Things like IBS, IBT, endogamy and shared ancestral location come in to play, but such expertise is way outside my square. These very interesting but rather deep technical discussions do help. Please avoid X as potential confusion with the male X chromosome adds a whole new ball game !.

I’m new to using wikitree and so I don’t know what you mean when you wrote (search g2g '41’).  I can see the “Search G2G” box But I should enter 41 in that box??  I’d like to read what you’re referring to.

I assume I was in effect using triangulation by doing all of the comparisons I did. 

I know what IBD is but I don’t know what IBT is.

I’ve done some reading on “sticky segments”.  As you said “very interesting” 

Identical By State by Type by Descent. Can't help beyond that as yet.

Just type 41 then look in the list that comes up. GS and I share 71 triangulated matches on C-17 which is a large number of small segment 'cousins' and is why this conversation arose. And yes, I think you are triangulating which is why I commented. As I see it, you can't do this in Ancestry. FamilyTree call it Matrix, MyHeritage call it triangulation and presumably GEDmatch can do it. I like the MyH option, but of course you miss out on all the people in other databases unless, like me, they have accidentally learned the advantages of uploading their test to other companies plus GEDmatch or GENi. I still have little idea which is the best option for looking at triangulation, let alone know how to interpret it. Good luck.

t sounds like my situation is similar to your situation.  You match GS with a total of 14.8 cM.  My sister matches TD with 16 cM on one segment.   But the number of smaller matches she (and I) have with TD’s known relatives is much less and the segment sizes are smaller (all between 4.0 and 5.0).  

For whatever reason, my sister, brother and I seem to have less non-distant matches on Ancestry than others who I have access to.  This could possibly explain why the number of matches my sister and I show is much less than the number of small segment matches you found.

If I understand correctly (per Frank Stanley ) in terms of “weight in numbers”  - there is.

From what you, Barry Smith, and Frank Stanley have written I’m tending to conclude  that even though there could well be actual connections with TD’s family, the likelihood of finding what those connections are is extremely small.  

A wider question is just how far back can sticky segments go ?. My BigY goes from Africa to the steppes of Russia back to Spain to become R1b then up to about Calais to become R-DF27 then across to GB. Then Wales as R-FT32960 then to Ireland to NZ, presently Terminating as R-FT34031.

My sticky segment is at the outer end of C-17 and appears to be stuck to FT34031 on the Y chromosome ?. atDNA puts me firmly in north central England, not all that far from Wales, but can a segment really stick for 10 or more generations with a coincidental Y link ?.
I am just learning this stuff, but found it interesting when I sorted my family finder list  by the surname I was looking for then ran those kits with geds through chromosome browser and segment detail then put those on a spreadsheet. I Did a bunch of them 7 at a time then sorted by Chromosome number and then sorted each chromosome by starting position. The pattern of overlap in the small segments and the large number of snips seemed like it should indicate kinship especially since they all had larger matches too. I was wondering if you ran enough related people that way using just the overlap if you could recreate the chromosome of the person you were related to?
Your last question is an interesting concept but difficult to understand why you would want to ?. ISOGG says we all have 6770 cM. Say on your mothers side, if you took an individual with the largest cM match and kept on adding the overlaps from other close maternal relatives you might get some of the way. A sibling or child has up to 50% from mum, but recombination would make the other 50% hard to rebuild and the maths might say impossible.

2 Answers

+2 votes
I think what cannot be ruled out is that those seven people all had a common ancestor who have them that segment, but that your match to that segment is false in the sense that it pulls some alleles from your maternal chromosome 9 and some from your paternal. If you could phase it with one of your parents then it would be more reliable.

Also, it is possible that it was a short segment that was widely spread through the community of your common ancestors. In that case, it would be “real”, inherited from one of your ancestors, but maybe not genetic proof that you received the segment from the ancestor you have identified.
by Barry Smith G2G6 Pilot (237k points)

Thank you for your response.

I found that I matched on the same area of chromosome #9 with 7 people all of whom are related to each other.  Likewise, except for one of them, my sister also matched - same area of the same chromosome.  I understand that what could appear as a match isn’t a real match due to alleles from one’s parents getting mixed in a “false” segment.  A real segment should all come from just my mother or just my father.  But does this happen often enough to explain that 35 of the 36 comparisons I made appeared to be matches when in fact they weren’t? Unfortunately my parents died some time ago so I can’t do any phasing.

I assume it’s possible that the 7 family members I’ve referred to plus my sister and myself could have all  inherited the same segment because everyone in the area they lived in had that segment.   I have no idea how one even thinks about proving or disproving this idea.

Thank you again.

Don't be fooled by the 36 comparisons. They are not independent comparisons. Once you know that you match person X on this little area of chromosome 9, then it immediately becomes *much* more likely that their child has that same segment. And in that case, you will match the child, and also the parent will match the child. For the less close relationships, the "*much* more likely" becomes less likely. But still, there are a lot of 5cM segments in the entire genome. If you look at just the group of close family members to X, there are probably quite a few segments larger than 5cM that they all share by being close relatives. So all of their comparisons will show matches on such segments. You just needed one little area of your genome to happen to match one of those segments they all share, and that may be what happened on chromosome 9.

I'd say you have good evidence that X's 5cM segment is real -- i.e., it doesn't jump between her maternal and paternal copies of chromosome 9, but is a real, contiguous segment on one chromosome that she got from one parent. It would still be possible that your own segment which matches actually jumps between your maternal and paternal chromosomes.

The fact that your sister matches on the same segment reduces the possibility of the latter. But on about 1/4 of your genome, you and your sister are "fully identical," meaning you have exactly the same maternal and paternal chromosomes. Your mother gave you and your sister the segment from the same one of your maternal grandparents, and your father gave you both the segment from the same one of your paternal grandparents. So there is still a decent chance your are fully identical on this part of chromosome 9, in which case both of your matches with X and her close relatives could be pulling from both maternal and paternal chromosomes. 

There is a simple way to check if you are fully identical. You and your sister could either upload to GEDmatch or else upload both kits to David Pike's tool "Search for Shared DNA Segments in Two Raw Data Files"

Their comparisons where show you "Fully Identical Regions" (FIRs). If you are half-identical in that part of chromosome 9, which is about a 50% chance, then that would be evidence that the segment is real.

But at that point, it could still be possible that the segment was very common in your ancestral population, in which case there might be other ancestors who could have sent it down to you, not just the one you have identified. But if the ancestral population has only a small amount of endogamy, then that's not much of a worry.

(BTW, I highly recommend identifying where you are half-identical and fully-identical with your sister. That information can be very useful for mapping the different parts of your genome to your four grandparents, which then makes it easier to cousin-match -- you already know which quarter of the tree to be looking in.)

Again, thank you for your response.

I think it’s fair to say that endogamy is a real possibility.  I’m of all Irish descent but the last of my ancestors came over around 1874.  My father’s family settled in NYC and my mother’s family settled in Philadelphia.  

The ancestry of the family I’ve been looking at is in County Leitrim. The counties I’ve identified so far as being the birthplaces of various ancestors are Longford (next to Leitrim), Limerick, Mayo, Galway, Tipperary and Donegal.  As you can see, my ancestors were from all over Ireland.

I understand only a limited amount about full and half identical regions but all of the people I’ve mentioned are on GEDmatch and that’s where I’ve done my comparisons.  I also have a 1st tier subscription and have used both the Segment Search and Multi Kit Analysis features to come up with some of the connections I’ve found.  Could you send me a link that you think gives a good explanation of how to use GEDmatch to help make the determination that the segment I’ve been looking at is “real”

Thanks, Carol

Not a link, but when you do the 1-1 autosomal comparison between your own kit and your sister's, check the second-to-last box from the bottom "Show only Full-Match (FIR) segments." If the segment you're looking at on chromosome 9 is not part of one of the resulting segments, but it is part of the segments where you match your sister when you don't check that box, then both of you have that segment on just one chromosome, so you'll have ruled out the maternal/paternal combo false match issue.

But if the segment is in one of the FIR matching segments, or else is not even shown as a matching segment between yourself and your sister when you run just the regular comparison, then it won't be ruled out.

I did as you suggested and there was no matching at all on cr9 when I ran the one-one comparison (with FIR clicked  ) between my sister and me.  However when I ran it between the brother and sister of the family in question, (with FIR clicked)  the segment that I’ve been looking at showed up in its entirety.  Does this indicate that both parents of the brother and sister could have matched on the same segment?

Generally, it could be that that brother and sister both got that segment from the same parent, say from their mother, or it could mean that the segment is matching you by bouncing between their maternal and paternal chromosomes -- because they have identical maternal and paternal chromosomes in that region, they could both be false-matching you in exactly the same way. However, because of the matches to their more distant relations, like the second cousin, it probably is a real segment they both inherited from one parent. .It sounds like yours is a real segment too. So it very well could have come from your common ancestor with that other family.

Jim Bartlett has been doing this for a long time and says that he feels pretty confident about most 5cM segments that triangulate and that survive the culling out of false matches as you have been doing are IBD:

But no one can give you definitive proof.

I’m confused.  When I did the one-to-one-comparison with my sister with FIR the fact that cr #9 didn’t appear on the list shows that the segment in question came from only one of my parents.  Yes??

But when I did the same with the brother and sister of the family in question and the segment in question DID show, that this added to the argument that the segment in question was not “REAL”.  The segment in question is between 31 million and 37 million and the brother and sister share between 28 million and 38 million as shown on the one-to-one list with FIR.

Thank you so much for your time with this.


+2 votes
Yes; It's possible you are related.  However, as pointed out the segment is small and could reflect a common ethnicity rather than family connection.  

What's not clear from the information provided is if the matching segments actually make sense.  

It's not enough that the matches are on the same Chromosome.  Are the the same location on the chromosome 9?

Also, do they make sense considering the closeness or distance of the relation?  

For example; do older generations, which would be expected to be more closely related have larger matches?

Likewise, do younger generations, which would not be as close to each other, have smaller matching segments?
by Andrew Ross G2G6 Mach 3 (31.1k points)

Thank you for your response.

Yes - all of the matches I’m referring to are in the same area on chromosome 9.  And nothing stands out as not making sense in terms of “closeness or distance of the relation”.

What I’m trying to get a handle on is whether the probability of my actually being  related to a person (who I’ll call X) who I match on a small segment increases as 1) the number of my family members who match X in the same area of the same chromosome increases and/or  2) the number of X’s family members who match me (same area of same chromosome) increases.

Related questions

+4 votes
2 answers
+8 votes
3 answers
+5 votes
1 answer
+5 votes
1 answer
1.4k views asked Mar 31, 2018 in The Tree House by Bennet George G2G6 Mach 1 (18.6k points)
+10 votes
1 answer
364 views asked Jan 13, 2018 in Genealogy Help by Barry Smith G2G6 Pilot (237k points)
+5 votes
3 answers
+4 votes
2 answers
+4 votes
2 answers
+6 votes
3 answers
428 views asked Sep 2, 2018 in WikiTree Help by Peter Mason G2G1 (1.5k points)

WikiTree  ~  About  ~  Help Help  ~  Search Person Search  ~  Surname:

disclaimer - terms - copyright