I have an impossible DNA match

Question

Ancestry tells me I share 1,892 cM across 56 segments

Unweighted shared DNA: 1,892 cM

Longest segment 109 cM

The results tell me that 100% of the time people sharing 1,892 cM have the following relationships Grandparent, Grandchild, Half Sibling, Aunt/uncle, Niece/nephew

I know the individual that I match couldn't possibly one of those, or any close relationship.  How likely is it that the result is just flat wrong.

I don't want to provide any details on a public board.

Comments

Sounds like the tester is not the person they say they are

maybe they attached the wrong name to the test to find info without revealing their identity

 My half sister is 1877 cM with 55 segments

so I am wonder how someone born 38 yrs before you  would even have that close of  a match

something is fishy about the tester IMO

---

To close this out, the DNA which I matched with did not come from the person I thought but another family member with the same given name.  To make things worse he is always called by a pet form of his middle name.  Once I made the connection, the result makes perfect sense.  A final thanks to everyone.

---

Thanks for the closure, John. Glad you found the explanation quickly.

Answer 1

The odds of this being a plain mistake are exceeding low.

Does this person share an identical X, Y or mtDNA with you?

How far apart in age are the two of you? If you check to see which relatives you have in common, do you recognize any of the names?

Answer 2

I agree with Jessica; the amount of sharing is highly unlikely to be incorrect by any significant amount. The only feasible error would be a complete mislabeling issue: ascribing test results to two different individuals.

The amount shared is right at the 25% level, which would equate to about 1,700cM. But the number of segments may tell a story. Grandparents and grandchildren (GP/GC) will, on average, have about half (well, about 45%) the number of shared segments as avuncular/materteral relatives (uncles, aunts, nieces, and nephews) and half-siblings (see this Henn et al. paper as one reference). Painting with a very broad brush, GP/GC will see a median of somewhere around 23 to 29 shared segments (meaningfully-sized segments of around 6cM or larger), while avuncular/materteral relatives and half-siblings will see a median of around 40 to 48 segments. Overlap can be seen at around 30 to 35 segments, and total counts can top 60.

AncestryDNA uses an algorithm called Underdog--a cute play on words since it's based on BEAGLE, a tool developed at the University of Washington for genotype calling and computational phasing--that if anything can tend to slightly underestimate the amount of shared DNA and introduce false breakpoints in some segments. Personally I'd consider the 56 segments reported by AncestryDNA likely to be spot-on with what you could expect for avuncular/materteral and half-sibling sharing. It's unlikely that Underdog would inflate GP/GC segments to that number. 

So barring a total mix-up in the labeling of two tests, I think that grandchild can be removed from the list of possible relationships here. And if generationally an aunt or uncle isn't feasible, that leaves a possible half-sibling, niece, or nephew.

AncestryDNA, of course, doesn't report matching on specific chromosomes, but as Jessica mentioned knowing the amount of xDNA sharing could provide a clue. A male half-sibling from your father would carry no matching xDNA, though a maternal half-sister would be expected to be a match the majority of the time, displaying at least around 15%, or about 30cM or more. You could find out by uploading both sets of test results to GEDmatch and running the comparison there.

The yDNA haplogroup could help if you're both male and he's a paternal half-sibling, or if he's the son of your father's brother. Otherwise the two of you wouldn't carry the same Y chromosome. There are ways to estimate your Y-haplogroup from your AncestryDNA test data, though it's worth noting that their autosomal tests after mid-2016 look at about twice the number of yDNA markers than the v1.0 prior to that, so the chances of a haplogroup prediction will, generally speaking, be better with newer tests from Ancestry.

Edited: Corrected by the alert and vigilant--which I wasn't--Simon Canning.  

Comments

Hi Edison,

I think that you are over-estimating the amount of maternal xDNA shared by full and half siblings. One example of two maternal half sisters share 112cM total and one example of a maternal half brother and sister share 101cM total.

When I look at the total xDNA for me, I share 42cM with my full brother and 93cM with my full sister. The same sister shares 67cM with my brother.

I also have three sisters that tested and they do have a total shared xDNA of 196cM.

This would seem to suggest that it would be more likely for full sisters to be the only ones that would share a total shared xDNA of 196cM. I am excluding parents from this as the relevant parent would show 196cM as well.

Cheers,

Simon

---

You're right, Simon. Way over-estimated. I should know better than to answer at 5:00 a.m. with no coffee yet. 

A paternal half-sister should, if the mothers are unrelated, match a male's xDNA not at all other than by chance (or identical by state). To try to dig myself out of that hole...

In males, the X chromosome is hemizygous, meaning that we have only one, but that it's joined at the pseudoautosomal regions with the the Y chromosome. So for males the raw data from most microarray autosomal tests display all alleles on the X chromosome as homozygous, i.e., AA, CC, GG, or TT (assuming no loci are no-calls), even though there is only the single, haploid X.

For matching purposes, our DTC testing and reporting companies look at half-identical regions, HIR, so any one of the two alleles from one kit corresponding to any one of the two alleles from the other kit would be considered a match at that locus.

The X chromosome undergoes recombination during meiosis in the mother, just like the autosomes. A male child gets his X chromosome entirely from his mother, but it might be anything ranging from inherited in whole from a single maternal grandparent, to made up of typically no more than five segments inherited from a combination of both maternal grandparents. A female child receives one haploid X intact from her father, and one haploid, recombined X from her mother.

In the case of paternal half-siblings, then, the daughter would carry the father's intact X chromosome, but the son would carry his mother's recombined X. The two should never have an actual X chromosome match unless the two mothers were closely related. Any reported matching would be in error, brought about by the fact that the male's X is hemizygous; that our microarray tests examine only between 15,000 and 30,000 loci of the X's approximately 156 million base pairs; and that by-chance results in the female's allele pairs showed enough HIR matching loci to be interpreted as a contiguous segment.

Consequently, yes, full-sibling sisters will always present as sharing 100% on the X chromosome because they each inherited the same intact X from their father. Male-male and male-female siblings will match only on the haploid, recombined X from the mother. Effectively the possibility exists for the rare but not impossible situations of zero matching between brother and sister up to a son and daughter both inheriting the same intact X from the mother; more probable is that male-female siblings will always display significant X chromosome sharing, but the amount will be highly variable.

Ah. Coffee...

<cough cough>

---

And just to double-back and clarify about the possibility of the results being a mistake...

The genotyping microarrays used for these tests sound like science fiction when you describe the way they work to someone, but in practice in the lab they're exceedingly simple. That's one reason the tests can be offered as inexpensively as they are.

AncestryDNA uses a chip believed to be manufactured by Quest Diagnostics, but they all operate--and are made--in a very similar manner. By way of comparison, the current GSA chip from Illumina has a reproducibility error rate of less than 0.1% (and that's out of over 654,000 SNPs) and a call rate of better than 99%. That last refers to the frequency with which the test can make a positive determination of the alleles present at any given locus. If the nucleotide values can't be determined, it's referred to as a "no-call" and a null value is reported for that location. In general, no-calls are simply ignored for matching purposes, just skipped over and assumed to be a match (HIR) if other SNPs fore and aft are matches.

Broadly speaking, at 1,800cM we're talking roughly 26% DNA sharing. So with a 0.1% error rate on microarray chips, you can see why we've discounted an error during the actual lab procedure itself. Someone's DNA really did match at approximately that level.

---

Hi again Edison,

I just wish to say that I would be very impressed with myself if I managed to be half as lucid and erudite as you without a single coffee.

I need at a minimum three coffees and at least that many hours to have elapsed before I can string a relatively coherent sentence together!

---

Thanks, Simon.    I try to embrace my mistakes because, well, that's how we learn. Considering my prolificacy in the errors and omissions department, by now I should be a lot smarter than I am. But yeah; what gray matter I have left needs to be jump-started with some morning caffeine. I shamble around like an extra in a zombie movie until the coffee starts to do its work.

Answer 3

I suppose you should have many matches in common with this match, and that can help figure things out. I would suggest trying to save as much information about the match, and matches in common, as possible. I have had, and many people I know of, matches getting skittish and deleting, hiding, or disappearing from results.

Comments

Those answers weren't what I was suspecting.  My match was born 38 years before me and before either of my parents were born so forget sibling.  He was born in Scotland and my grandparents were all born in the US and I'm sure I know of all my parent's siblings so I would rule out uncle unless my parents had siblings they weren't aware of.  My Grandparents were born in IA, SD, KS and my contact was born in Scotland, immigrated to NY and died in CT so I don't see how he could be a newly found fifth grandparent.  Have I left anything out?

This is a new autosomal test so I don't think I have any xDNA or haplogroup information.

my next closest match is 267 cM over 13 segments so this is a real outlier.  I can't trace his tree very far so there are no common names.

Thanks for your help, wish Ancestry did as much.

---

Hi John,

Based on your birth year range from your profile (1940s), if your match was born 38 years before you, then the match was born in 1911 or earlier. Are you conversing with the actual match (aged 109 years or older), or a proxy for a deceased match?

---

My father died in 2007, and through DNA a few years ago, I discovered an unknown (still living) half-sibling of his. He never knew. The answer for me was my dad's father was not a Lowery, but a different , man I had never heard of, who lived a few hundred feet from my grandma. Some secrets people never reveal.

---

I'm not conversing with anyone yet.  The match is deceased.  I don't know who the proxy is but I'm sure I know them and they know me and would be as shocked by this result as I am. Thus my caution and wish to rule out some testing error.

---

just wondering if you ever had a conversation with this half-sibling and if so, how it went.

---

Since you've not been conversing with the contact person for the match, is the information that you have about the match from the profile of an individual in an Ancestry family tree that is linked to the match?

I ask this because I've found multiple occurrences in my AncestryDNA match results in which a match or proxy has associated the test info to the wrong individual in a tree (usually incorrectly connected to a parent of the match, which distorts the likely generational difference between the match and I).

That may not be the case here, but something else to consider.

---

Some matches of mine are connected to a tree, whose root person is incorrect, meaning not connected to said match in tree.

---

Rick has a very valid point. Other than two test kits being mislabeled somewhere along the way, the other place for error would be incorrect vital information for the match.

What we can tell from that 1,892cM and 56 segments, though, is that there is very little wiggle room for other possible relationships. It probably isn't a hitherto unknown biological grandparent, even if the year of birth is correct, given the distribution of autosomal segments I described earlier. Should be closer to 30, not 50, if it were a grandparent. And there really are no reasonably feasible relationships if the person was born a generation before you (and by that I'm excluding a large gap between half-siblings; I personally have a gap of almost 30 years, but 38--while possible for a father--would be stretching it) that would explain that size match. Not to be indelicate, but even incest between a great-grandmother and her uncle would not account for a great aunt/uncle of yours having that ~25% sharing with you.

Looking at Rick's possibility that the birth year and the tree information is off for this match, the one thing I can think of would be double 1st cousins. If your grandparents each had (probably older) siblings who married each other, then there would have been pedigree collapse at the great-grandparental level: instead of having 8 different g-grandparents, you have  4...which your WikiTree ancestry, of course, indicates not at all.

But double 1st cousins would present roughly the same percent sharing and segmentation as half-siblings. Whomever those DNA results belong to, however, she or he is going to be related to you in one of the ways we've discussed. At the next level of genetic relationship, 12.5% sharing, there are a larger number of possibilities to explain the DNA. There simply aren't many at ~25%.

---

Edison, (and Rick)

Just before I saw your comment, I realized that if the matching DNA actually came from the grandson of the identified source, then the result makes perfect sense, in fact has been known all along.  Barring actually communicating with the proxy, I'm going to assume that is the case since it makes more sense than anything else.

Again thanks for everyone's help.

---

That would certainly 'splain it...though still a larger than expected number of segments.

The only thing I hate about genetic genealogy mysteries are the ones that go unresolved. Keep us informed?

---

John - If the questio. Re talking to half-sib was dor me, I sent snailmail letter explaining why I thought the lady was my half-aunt, with pictures if ny dad etc. She emailed me & was a little upset I guess, & I sent more info on my matches, map showing distance between her dad & my grandmother... in between the snailmail & emails to me I contacted her granddaughter to ask her to test... I also told my half-aunt what kind  of  of man my dad was etc.  In the end my half- aunt was OK with her granddaughter testing, & it all fit. Everyone's experience in these situations is different.i hope it works out amicanly for you John & your new found relative.