PDA

View Full Version : Triangulated matches and variance among family members



BalkanKiwi
03-27-2021, 07:04 AM
Friends,

I've encountered a puzzle that I haven't come across before. I've stumbled upon (I was looking at something else and came across this instead) a triangulated group of Dutch matches on my grandmother's side on MyHeritage. What's interesting about this group is that a number of my cousins (including a first and second cousin) are also involved, which helps narrow down the side of the family.

Its a triangulated group of 7 Dutch matches (including a mother and son), and its an 8cM segment with 3584 SNPs. All of them are from the Netherlands, with numerous Dutch surnames in their trees. This is on an English branch of the family.

The issues I have are:

1. I match and triangulate with all 7.
2. My grandmother matches with 5 of them, but only triangulates with 4 of them.
3. My first and second cousins only match 1-2 of them.

Are these possibly a bunch of false matches due to the low segment size? Or is it likely a real triangulated group, and something like SNP variance among testing companies is causing the differences?

If someone has come across such a situation before, any input would be great.

Tolan
03-27-2021, 07:44 AM
I have serious doubts about the reliability of Myheritage matches.
I have lots of inconsistencies like this.

I also have for example another inconsistency:
I sent a phased kit from my mom, and she has matches that I don't have.
Which is impossible since the phased kit comes from me!


I think it's because of their imputation. Imputation involves guessing the results of untested markers from known neighboring markers.
It is an aberration for me, but it seems that it is scientifically proven.
But 50% of scientific research published in journals are biased and are contradicted by subsequent studies!>:(

Certainly that many segments are therefore probably not recognized as identical

Pylsteen
03-27-2021, 07:57 AM
I also have some which triangulate with e.g. my grand uncle but where my own grand parent only has half of them. Because of the triangulations, I consider the matches as probably real, although likely far back in time. In some cases the segments may be close to the treshold, then slight variations in the matching locations might play a role too .

FionnSneachta
03-28-2021, 02:00 PM
I also have for example another inconsistency:
I sent a phased kit from my mom, and she has matches that I don't have.
Which is impossible since the phased kit comes from me!

I might be misunderstanding but I phased my kit against my mum's on GEDmatch. My mum shares 76.1 cM with her 5th highest match on GEDmatch. I don't match this person at all on GEDmatch. However, after I phased my kit against my mum on GEDmatch, this phased kit shares 72.9 cM with this match that I don't match with my own kit. Should that not happen?

geebee
03-30-2021, 12:10 AM
I have serious doubts about the reliability of Myheritage matches.
I have lots of inconsistencies like this.

I also have for example another inconsistency:
I sent a phased kit from my mom, and she has matches that I don't have.
Which is impossible since the phased kit comes from me!


I think it's because of their imputation. Imputation involves guessing the results of untested markers from known neighboring markers.
It is an aberration for me, but it seems that it is scientifically proven.
But 50% of scientific research published in journals are biased and are contradicted by subsequent studies!>:(

Certainly that many segments are therefore probably not recognized as identical

I'm a little confused. You say, "I sent a phased kit from my mom, and she has matches that I don't have."

I was not aware that MyHeritage has any ability to recognize a kit as having been phased, especially if you had to convert the kit into some format that could be uploaded to them. I also don't know what kind of phasing MyHeritage uses as part of its "ancestry analysis" -- if any. It's possible that the only phasing they use is statistical phasing, and in that case they would have attempted that on whatever file you sent them (already phased or not).

In any case, your mother should have many matches you don't, since she is likely to have matches in the half of her DNA that she did not pass on to you. At any given location in her genome, she either passed on to you a segment that came from your father or a segment that came from her mother. Occasionally, a segment could be partly from one and partly from the other -- but usually only for segments which are fairly large.

If you inherited a chunk of DNA that came from your mother's father, then you won't share a match that your mother inherited from her mother if it's in that same location. And of course the reverse is true: if the DNA you inherited from your mother at a certain location came from her mother, you won't share matches that are on your mother's father's side in that same location.

I see this also the time with my father at 23andMe. He has matches that he may share with one or more of my siblings in a given region, but not with me; or the other way around. All that means is that my siblings and I didn't inherit the same paternal DNA at that location.

geebee
03-30-2021, 12:30 AM
I might be misunderstanding but I phased my kit against my mum's on GEDmatch. My mum shares 76.1 cM with her 5th highest match on GEDmatch. I don't match this person at all on GEDmatch. However, after I phased my kit against my mum on GEDmatch, this phased kit shares 72.9 cM with this match that I don't match with my own kit. Should that not happen?

There is no reason you should expect to share all of your mother's matches, since she only passed on half of her DNA to you. This is even possible for a very large segment, or for several segments. I've seen instances in which full siblings share no DNA at all on a given chromosome, meaning that they inherited from opposite grandparents on both sides! This isn't common, but it happens. (You certainly would not expect it to happen at more than a chromosome or two, however.)

The strange thing here, though, is that it really shouldn't matter whether you use a phased or unphased kit with a match this large. To be a match -- or technically, a half match -- it's only necessary that you and another person share at least one SNP value at certain sequence of SNPs. Typically, a few mismatches are allowed, though there may be limits as to how close together they're allowed to be (where both SNPs are homogenous but opposite for you and the other person). Also, depending on the algorithm, no calls may be ignored or not ignored.

The bottom line is, phasing isn't required for matching purposes, because matching usually does take both sides into account -- and matching only has to be a single letter at each SNP location. (So for example, CG would be a half match to CT, because both have "C". GG would not be a match to TT, but a single instance in what otherwise would be a matching segment would probably be ignored.)

peloponnesian
03-30-2021, 08:21 AM
I've seen instances in which full siblings share no DNA at all on a given chromosome, meaning that they inherited from opposite grandparents on both sides! This isn't common, but it happens. (You certainly would not expect it to happen at more than a chromosome or two, however.)

Me and my sister:

https://i.imgur.com/7aKXwhc.png

FionnSneachta
03-30-2021, 09:09 AM
There is no reason you should expect to share all of your mother's matches, since she only passed on half of her DNA to you. This is even possible for a very large segment, or for several segments. I've seen instances in which full siblings share no DNA at all on a given chromosome, meaning that they inherited from opposite grandparents on both sides! This isn't common, but it happens. (You certainly would not expect it to happen at more than a chromosome or two, however.)

The strange thing here, though, is that it really shouldn't matter whether you use a phased or unphased kit with a match this large. To be a match -- or technically, a half match -- it's only necessary that you and another person share at least one SNP value at certain sequence of SNPs. Typically, a few mismatches are allowed, though there may be limits as to how close together they're allowed to be (where both SNPs are homogenous but opposite for you and the other person). Also, depending on the algorithm, no calls may be ignored or not ignored.

The bottom line is, phasing isn't required for matching purposes, because matching usually does take both sides into account -- and matching only has to be a single letter at each SNP location. (So for example, CG would be a half match to CT, because both have "C". GG would not be a match to TT, but a single instance in what otherwise would be a matching segment would probably be ignored.)

I understand that, of course, my mum will have matches that I don't have. I was just wondering does it make sense that I can match someone after phasing with a parent if I don't match them at all with my regular non-phased kit? If that does make sense and is unsurprising, then that's fine.

C J Wyatt III
03-30-2021, 04:12 PM
Me and my sister:

Thanks, I am surprised at the extreme results.

I am curious which test and version that you two used?

peloponnesian
03-30-2021, 04:51 PM
Thanks, I am surprised at the extreme results.

I am curious which test and version that you two used?

It's from the latest MyHeritage chip uploaded to Gedmatch. I wonder how it would look if we both did whole genome testing.

C J Wyatt III
04-01-2021, 01:47 AM
It's from the latest MyHeritage chip uploaded to Gedmatch. I wonder how it would look if we both did whole genome testing.

What I think would be interesting would to show the chromosome #14 comparison using full resolution. I am predicting that you will see some yellow mixed in with the green.

peloponnesian
04-01-2021, 10:18 AM
What I think would be interesting would to show the chromosome #14 comparison using full resolution. I am predicting that you will see some yellow mixed in with the green.

Yeah, with the full resolution there were 5 tiny slivers of yellow.

C J Wyatt III
04-01-2021, 05:12 PM
Yeah, with the full resolution there were 5 tiny slivers of yellow.

I appreciate it. I would have expected more.

geebee
04-06-2021, 01:46 PM
Me and my sister:

https://i.imgur.com/7aKXwhc.png

Here's a comparison between my eldest brother and me at GEDmatch, first at lower resolution:

44173

Here's a portion of the same comparison at full resolution:

44174

You can see that somewhere after genomic position 20 million, there are a number of "yellow slivers". What do they mean? Actually, not much. According to the key, yellow is for individual markers that have half matching base pairs. That simply means that at a particular SNP two people have only one common value -- for example, if one has "CT" and the other has "GT", there is a half match. If it's just a few individual markers, it may just be happenstance. To determine a genuine region of half matching, the half matching has to continue over a significant number of SNPs.

Now, here's the same comparison done at 23andMe:

44175

What this tells me is that my brother and I essentially match on both our maternal and paternal copies of chromosome 16. There is a very small region at which 23andMe says there is half matching only -- meaning, we only match on either our paternal or maternal copy of chromosome 16 in this location.

To look at it numerically, 23andMe says we have half matching from genomic position 1 to genomic position 90,149,922 -- or for 134.02 cM and 17,421 SNPs. This is pretty much the entire chromosome. We also have complete matching from 97,354 to 22,097,738 (42.98 cM and 5,155 SNPs) and from 22,239,666 to 90,149,922 (90.99 cM and 12,244 SNPs).

If you add up the SNPs with complete or full matching, you get 17,299 -- or just 122 SNPs that have only half matching. This is actually a fairly small number.

If you look at the numbers at GEDmatch, matching segments are reported from 105,320-18,165,043 (39.7 cM and 3,585 SNPs); 18,807,311-31,957,368 (18.3 cM and 1,602 SNPs); and 46,640,695-90,149,922 (74.0 cM and 8,393 SNPs). The sum of the SNPs is 13,580, which is significantly differ from what 23andMe is reporting. I'm not sure of the reason for that. It may be related to the fact that my brother and I tested on different chip versions (v2/v3 for me and v4 for my brother). It's possible that 23andMe is imputing some of my brother's SNPs, which would actually not be that unreasonable -- especially if they also take our father's results into consideration. (They probably don't, however.)

My point is that you shouldn't put much stock in "slivers" of yellow. Between parent and child, the inherited segments from one or both grandparents will generally be quite large. That's because there will be very few crossovers on each chromosome. There might be none, for some chromosomes.

There will also be instances in which a particular SNP has a miscall or a no call for at least one person being compared. Some comparison algorithms will ignore a certain number of these, especially when they're far apart. Some may not.

All that a "sliver" of half matching necessarily means is that at a particular SNP, the two people being compared just "happened" to have a single common base (refer to the first paragraph of this post). That will happen sometimes with people who aren't even related in any genealogical time frame. It certainly does not suggest that many crossovers occurring very closely together!

EDIT:

I didn't comment on the chromosome 15 comparison, but the bottom (black) bar shows "no matching" for pretty much the entire chromosome. Although there's a lot of yellow "half matching" in the top bar, it doesn't really mean anything. GEDmatch is telling you that, in this case, you and your sister inherited exactly opposite DNA from both parents -- that is, DNA from the opposite grandparent on each side.

My second brother and I have only half matching on a very small portion of chromosome 18. Specifically, from 73,287,109-78,015,180 (11.34 cM and 1,253 SNPs). So we inherited everything except for the 11.34 cM region from opposite grandparents on each side.

It may not happen very often, but sometimes there will be a chromosome on which even two full siblings do not match at all. I believe this is most likely to occur when each sibling inherits an unrecombined copy of that chromosome from the opposite grandparent on each side.

"Unrecombined" chromosomes are transmitted more often than one might think, perhaps a few times in each meiosis.

For example, this occurred at least once with each of my father's five tested grandchildren. They are not listed by age order, but ##1 and 2 are a son's daughter and son; ##3 and 4 are a daughter's daughter and son; and #5 is a son's daughter.

No DNA shared with my father on these chromosomes:


Grandchild #1 - 17, 19, 21, X*

Grandchild #2 - 4, 9, 12, 13, 18, 19, 21, X*

Grandchild #3 - 9, 16, 22

Grandchild #4 -

Grandchild #5 - 7, 15, 17, 18, 22, X*

Notice that only Grandchild #4 inherited DNA from my father on at least part of each chromosome. Also, ##3 and 4 are my sister's children, and therefore are the only ones to receive DNA from our father's X chromosome.

For what it's worth, Grandchild #4 shares the most DNA with my father at 27.97%, and Grandchild #2 shares the least at 18.46%.

Also, at least two grandchildren received chromosomes which appear to have come entirely from my father. Specifically, Grandchild ##1's copies of chromosomes 16 & 20 and Grandchild #5's copy of chromosome 9 all appear to (half) match their grandfather across their full length.

This suggests that in both cases, these two received an unrecombined copy of their respective father's paternal copy of these chromosomes.

peloponnesian
04-06-2021, 04:31 PM
.

Thanks for your explainer, appreciate it. It's one of those cases where I can relay the concept relatively easily to my parents.

23abc
04-06-2021, 04:51 PM
Thanks, I am surprised at the extreme results.

I am curious which test and version that you two used?

My two sons inherited a very different amount of DNA from each grandparent. 23andMe says they share around 37% while AncestryDNA says they share 41-46% (gives a range)... not sure how there can be such a big difference. Perhaps because one includes the X chromosome and the other doesn't?

https://imgur.com/uXQcbU2.png

evon
04-07-2021, 09:21 AM
Friends,

I've encountered a puzzle that I haven't come across before. I've stumbled upon (I was looking at something else and came across this instead) a triangulated group of Dutch matches on my grandmother's side on MyHeritage. What's interesting about this group is that a number of my cousins (including a first and second cousin) are also involved, which helps narrow down the side of the family.

Its a triangulated group of 7 Dutch matches (including a mother and son), and its an 8cM segment with 3584 SNPs. All of them are from the Netherlands, with numerous Dutch surnames in their trees. This is on an English branch of the family.

The issues I have are:

1. I match and triangulate with all 7.
2. My grandmother matches with 5 of them, but only triangulates with 4 of them.
3. My first and second cousins only match 1-2 of them.

Are these possibly a bunch of false matches due to the low segment size? Or is it likely a real triangulated group, and something like SNP variance among testing companies is causing the differences?

If someone has come across such a situation before, any input would be great.

Matches do not always triangulate, even when they should, it is not a problem that is limited to matches alone, or to any one DNA company. I have encountered issues with 23andme (including their Ancestry composition), FTDNA and Myheritage. I am currently working on one such issue using DNAPainter with data from Myheritage that is driving me nuts :P My only consolidation, at the moment, is that I am working on DNA segments that are very distinct and so they are easier to label correctly based on the profile of the matches (Finnic segments in the Nordic Romani gene-pool), but I work very hard to rule out any human errors that I may commit in the process.

Saetro
04-07-2021, 11:08 PM
Friends,

I've encountered a puzzle that I haven't come across before. I've stumbled upon (I was looking at something else and came across this instead) a triangulated group of Dutch matches on my grandmother's side on MyHeritage. What's interesting about this group is that a number of my cousins (including a first and second cousin) are also involved, which helps narrow down the side of the family.

Its a triangulated group of 7 Dutch matches (including a mother and son), and its an 8cM segment with 3584 SNPs. All of them are from the Netherlands, with numerous Dutch surnames in their trees. This is on an English branch of the family.

The issues I have are:

1. I match and triangulate with all 7.
2. My grandmother matches with 5 of them, but only triangulates with 4 of them.
3. My first and second cousins only match 1-2 of them.

Are these possibly a bunch of false matches due to the low segment size? Or is it likely a real triangulated group, and something like SNP variance among testing companies is causing the differences?

If someone has come across such a situation before, any input would be great.

When I have come up against such matches that a parent does NOT have, the only ones that make sense are when there has been a misread in the middle of a parent's segment, causing it to be scored as two smaller ones below the threshold.
Otherwise, I tend to count any segment of mine with no parental match also, as false.
But I don't yet have enough examples from MyHeritage to properly understand what is going on there.

All instances so far of me having tiny matches with lots of SNPs, have matches with no possible CA for about 4 centuries, and where the match's ancestors came from a tiny genepool. So I have inferred effective endogamy on their side. Successive discussions about that particular genepool has found other experienced DNA genealogists have reached the same opinion.
But I can't quite see how that applies to your situation.
Except the CA being back a long way.

BalkanKiwi
04-08-2021, 04:41 AM
When I have come up against such matches that a parent does NOT have, the only ones that make sense are when there has been a misread in the middle of a parent's segment, causing it to be scored as two smaller ones below the threshold.
Otherwise, I tend to count any segment of mine with no parental match also, as false.
But I don't yet have enough examples from MyHeritage to properly understand what is going on there.

All instances so far of me having tiny matches with lots of SNPs, have matches with no possible CA for about 4 centuries, and where the match's ancestors came from a tiny genepool. So I have inferred effective endogamy on their side. Successive discussions about that particular genepool has found other experienced DNA genealogists have reached the same opinion.
But I can't quite see how that applies to your situation.
Except the CA being back a long way.

That's interesting. I also tend to ignore matches who don't match a parent or a grandparent. I have a similar situation with a bunch of Finnish matches. My grandmother matches most of them, the first cousin (the same one involved in this thread discussion) matches a few of them on MH, but not all. I've found 46 kits on GEDMatch (including a number of the Finnish MH matches) who mostly triangulate. Even zoomed out at 30% I couldn't fit them all in. I'm at the top and my grandmother is second. Most of the matches are 7-8cM, but we don't match all of them (I match and triangulate with 25 for example).

I'm going to assume we probably share a common ancestor, even more so now that 23andMe assigned me Finnish in their recent update. I also have seven 100% Finnish matches on 23andMe, and a large number on FTDNA.

https://i.imgur.com/7UrFBET.jpg

lana6765
04-08-2021, 10:07 PM
That's interesting. I also tend to ignore matches who don't match a parent or a grandparent. I have a similar situation with a bunch of Finnish matches. My grandmother matches most of them, the first cousin (the same one involved in this thread discussion) matches a few of them on MH, but not all. I've found 46 kits on GEDMatch (including a number of the Finnish MH matches) who mostly triangulate. Even zoomed out at 30% I couldn't fit them all in. I'm at the top and my grandmother is second. Most of the matches are 7-8cM, but we don't match all of them (I match and triangulate with 25 for example).

I'm going to assume we probably share a common ancestor, even more so now that 23andMe assigned me Finnish in their recent update. I also have seven 100% Finnish matches on 23andMe, and a large number on FTDNA.

Ah the Finnish matches!

I have so many matches from Norway, Sweden and Finland that I am starting to wonder whether the small amounts of 'Finnish' and 'Norwegian' I got with FTDNA and my initial AncestryDNA estimate were in fact based in some truth.

I'm talking individual segments 15-23 cM long and overall match quality of >30 cM.

I don't know how we are related, but it's not impossible, even from someone with 'only' English and Irish ancestry.

Sadly, both parents have decided not to do DNA tests.

Some of the Finns triangulate with some of the Swedes and some just match each other and me. It's like we're all one big lost family :D

Riverman
04-08-2021, 10:41 PM
My Heritage is definitely not the most reliable site for comparisons, but mistakes being done by all companies and Gedmatch too. In some cases I can even tell which company did wrong for a match, and sometimes its a miss or false match in X, the next time in Y. The most reliable matches overall seem to come from Ancestry DNA, but at the cost of many matches getting lost in the selection process, which seem to be real based on the data I could look at. So there is no perfect offer out there yet, just like with ethnicity and regional estimates. Some are better most of the time, but none all the time.

There is definitely an issue on My Heritage with small scale SNP variation causing missing matches, false negatives and surplus matches, false positives. Minor alterations in the data can cause both, which I could see by comparing clusters of matches from different uploads. The differences can be quite big, which tells you a lot.

Concerning Scandinavian and Finnish matches, I came to the conclusion that many are real and based on continental immigrants, which spread their genes quite successfully in fairly small communities, leaving behind a lot of descendants. I could trace back only a small minority of these, but it turns out its always going from continental Europe, even from small, less mobile communities, to Scandinavia and Finland. A recently detected, particularly interesting case was about German mercenaries in Finland. My match had two of these, independently, on different branches of her family tree. And it was many generations back, in the 18th century. I mean how many people even have a tree which goes that far back? In another case the possible last common ancestor was from the end of the 15th century and both this ancestor and the other triangulated matches point to the same ancestral source region. This just means to me that in some cases even very long segments can survive for many generations. And some of these matches or segments appear on all platforms, so that it shouldn't be a glitch coming just from the failure of one company.

But yes, especially with My Heritage there is this feeling that matches don't count as much as on some other sites, because there are just too many false positives and low level matches included. But then again, some of these prove to be right, and they being missed on the other platforms completely. You can't get it all, because there is just the choice between too many false positives and too many (missing!) false negatives. I don't even think that's just a technical issue, its the way the DNA being inherited which includes a factor of randomness and uncertainty for small matches in particular.

BalkanKiwi
04-09-2021, 10:05 AM
Ah the Finnish matches!

I have so many matches from Norway, Sweden and Finland that I am starting to wonder whether the small amounts of 'Finnish' and 'Norwegian' I got with FTDNA and my initial AncestryDNA estimate were in fact based in some truth.

I'm talking individual segments 15-23 cM long and overall match quality of >30 cM.

I don't know how we are related, but it's not impossible, even from someone with 'only' English and Irish ancestry.

Sadly, both parents have decided not to do DNA tests.

Some of the Finns triangulate with some of the Swedes and some just match each other and me. It's like we're all one big lost family :D

I've been wondering how common it is for those with English, Irish, Scottish etc ancestry to locate a reasonable number of triangulated Scandinavian/Finnish matches. I've had some success (https://anthrogenica.com/showthread.php?23458-Mapping-out-common-Genetic-Groups-among-clusters) in mapping out Norwegian and Danish matches.

If you've managed to find a number of triangulation's, its probably a good bet they are real matches.

lana6765
04-09-2021, 11:12 AM
I've been wondering how common it is for those with English, Irish, Scottish etc ancestry to locate a reasonable number of triangulated Scandinavian/Finnish matches. I've had some success (https://anthrogenica.com/showthread.php?23458-Mapping-out-common-Genetic-Groups-among-clusters) in mapping out Norwegian and Danish matches.

If you've managed to find a number of triangulation's, its probably a good bet they are real matches.

I don’t know how common it is for the population as a whole, but for reference, my ancestry comes mainly from South and East England, especially London and Essex. I’m also around a 3rd West Irish, mainly from in and around Westport, County Mayo, and a teensy bit Liverpudlian.

Also a lot of my family were involved in sailing and sailed as far as France, Australia and the Americas.

lana6765
04-09-2021, 11:29 AM
Concerning Scandinavian and Finnish matches, I came to the conclusion that many are real and based on continental immigrants, which spread their genes quite successfully in fairly small communities, leaving behind a lot of descendants. I could trace back only a small minority of these, but it turns out its always going from continental Europe, even from small, less mobile communities, to Scandinavia and Finland. A recently detected, particularly interesting case was about German mercenaries in Finland. My match had two of these, independently, on different branches of her family tree. And it was many generations back, in the 18th century. I mean how many people even have a tree which goes that far back? In another case the possible last common ancestor was from the end of the 15th century and both this ancestor and the other triangulated matches point to the same ancestral source region.

I may never solve how we’re related. We might not be related over the last 200 years. And what makes it difficult is that tests can miss minor ancestries. The matches may appear to be 100% Scandinavian or 98% Finnish, 2% Inuit, but they may not be.

My DNA pulls towards both Finland and towards Southern France and Iberia. I also tend to get a small amount of Southern European on these tests. I think I have relatively high Western Mediterranean considering my background and also some small Siberian and East European component, going by G25. So the matches could be related to... pretty much anywhere in Europe. I have some more distant matches from Italy and Eastern Europe too.

Riverman
04-09-2021, 12:41 PM
I may never solve how we’re related. We might not be related over the last 200 years. And what makes it difficult is that tests can miss minor ancestries.

That's indeed the case for many of my matches with proven Central European ancestry. The minority ancestry disappears in the estimate, but its there. It always depends on what kind of ancestries we're talking about, because some are very close, others not. Those which are close eat each other up on most platforms. Just compare with some matches with a good tree, even if they are low for your. I guess you will find some matches which are 25 percent of a different ethnicity which got swallowed or reduced to trace levels. That happens even on Ancestry DNA and My Heritage is unreliable with its ethnicity estimates anyway. The genetic groups become totally unreliable below 25 percent too between related groups which are hard to differentiate.

BalkanKiwi
04-10-2021, 06:17 AM
In an interesting turn of events, 1-2 of the Dutch matches I mentioned in the first post, are indirectly related to the Finnish cluster. Summarised:

1. A few of the Dutch matches triangulate with two Swedes on the Finnish segment (one has 18% Finnish, the other 68% Finnish).
2. The Dutch matches don't triangulate with any of the Finns directly.
3. The Swede with 18% Finnish triangulates some of the Finns.

So basically, there is one Swede acting as the middle person, who links the Dutch with the Finns. Its possible they are false positives, but its plausible that we all share a common Finn ancestor in some way. Given the size of the segment, its certainly a fair way back in time.

Riverman
04-10-2021, 07:39 AM
In an interesting turn of events, 1-2 of the Dutch matches I mentioned in the first post, are indirectly related to the Finnish cluster. Summarised:

1. A few of the Dutch matches triangulate with two Swedes on the Finnish segment (one has 18% Finnish, the other 68% Finnish).
2. The Dutch matches don't triangulate with any of the Finns directly.
3. The Swede with 18% Finnish triangulates some of the Finns.

So basically, there is one Swede acting as the middle person, who links the Dutch with the Finns. Its possible they are false positives, but its plausible that we all share a common Finn ancestor in some way. Given the size of the segment, its certainly a fair way back in time.

There are also pile-up and "sticky segment" regions on the DNA I guess. It seems these stay intact longer and sometimes even for selective reasons:
https://isogg.org/wiki/Identical_by_descent

But that's a sensitive issue, because the definition, inclusion or exclusion of such segments is probably not always that easy to decide.

I came across the fact that I have most large matches which are truly distant on segments which I inherited exclusively from one specific ancestor older than great-great-grandparent. So for a lot of the old and long segments, the limiting factor is my own genome, how much I got from one specific ancestor. I could test that with known cousins, from which I got very big chunks of DNA, making up a large fraction of the chromosome. And voila, I got very high other matches on the very same segments too, even though some are only quite distantly related, with a common ancestor not found and most likely having lived many centuries ago. They just have happened to inherit similar sized chunks of the same common ancestor. And I can even see that in detail for those which have had their parents and siblings, some even their children tested. Because from one generation to the next, even though the segment is more than 30 cM and probably 400 years old, it can be broken up - or not. Its pure chance. Especially single segments are unreliable even if they are fairly large imho. You never know how old it is, because by chance it can be "kept alive" for many more generations than what is the statistical average.

BalkanKiwi
04-10-2021, 11:10 AM
There are also pile-up and "sticky segment" regions on the DNA I guess. It seems these stay intact longer and sometimes even for selective reasons:
https://isogg.org/wiki/Identical_by_descent

But that's a sensitive issue, because the definition, inclusion or exclusion of such segments is probably not always that easy to decide.

I came across the fact that I have most large matches which are truly distant on segments which I inherited exclusively from one specific ancestor older than great-great-grandparent. So for a lot of the old and long segments, the limiting factor is my own genome, how much I got from one specific ancestor. I could test that with known cousins, from which I got very big chunks of DNA, making up a large fraction of the chromosome. And voila, I got very high other matches on the very same segments too, even though some are only quite distantly related, with a common ancestor not found and most likely having lived many centuries ago. They just have happened to inherit similar sized chunks of the same common ancestor. And I can even see that in detail for those which have had their parents and siblings, some even their children tested. Because from one generation to the next, even though the segment is more than 30 cM and probably 400 years old, it can be broken up - or not. Its pure chance. Especially single segments are unreliable even if they are fairly large imho. You never know how old it is, because by chance it can be "kept alive" for many more generations than what is the statistical average.

Indeed regarding pile ups. I normally quickly refer to here to see if clusters are within these regions. The Finns match on chromosome 10 but are away from the suggested pile up region.

http://www.khazaria.com/genetics/excess-ibd-regions.html

lana6765
04-14-2021, 10:41 PM
Indeed regarding pile ups. I normally quickly refer to here to see if clusters are within these regions. The Finns match on chromosome 10 but are away from the suggested pile up region.

http://www.khazaria.com/genetics/excess-ibd-regions.html

For what it’s worth, they’re all likely real, but the question here is time.

"In a study of a European subset of the Population Reference Sample (POPRES) dataset it was estimated that for the most part IBD blocks longer than 4 cM come from 500 to 1,500 years ago, and blocks longer than 10 cM are within the last 500 years."

I mean think how many ancestors we all have from the 1500 generation. Assuming everyone has children at 33.3 years and you were born in 2000: 65536. Okay maybe a bit less than that as you might have 63 4x great grandparents. But it would still be a pretty huge number even with distant half cousins marrying.

cvolt
05-03-2021, 10:38 AM
I pretty much only get Swedish or Finnish matches now. I should have some because I am half Swedish, but it definitely seems way way overshot compared to 23andme or Ancestry matches. Also, I think their Scandinavian is way too close to other NW pops. On 23andme my whole family has the right amount of Swedish, German, etc... so it's definitely possible to differentiate the ethnicities.

BalkanKiwi
05-16-2021, 04:32 AM
I've recently come across a similar (but more straight forward) situation with Finnish and my father. 23andMe gives me 0.8% Finnish (this stays at 70% confidence) on a small segment on chromosome 9. Using FTDNA I've worked out this comes from my father. My father and I both triangulate with a reasonable number of Finns along this exact segment on chromosome 9 (on both MyHeritage and FTDNA), most of whom are close to, if not, 100% Finnish. Some of them my father and I both triangulate with, some we match individually but still in the same location. On 23andMe I have a five 100% Finnish matches (2 of whom are predicted 5th cousins, which is unrealistic). One of the 23andMe matches also matches along the chromosome 9 segment.

Regarding where this Finnish ancestry came from, I assume its related to the Polish/Eastern Euro side.