Can autosomal DNA confirm to the common ancestors?

Question

The help page for autosomal DNA confirmation states:

If your match is a first cousin on your father's side you can confirm your relationship to your father and his relationships to both his parents. If your match is a first cousin on your mother's side you can confirm your relationship to your mother and her relationships to both her parents.

Wouldn't it be true that the confirmation is to at least one of the common ancestors, but not necessarily to both of them? If the DNA in common were inherited entirely from the common female ancestor, then the genetic father might not have been her husband. Also, but less likely, the common male ancestor could have fathered a child with another woman, and then adopted the child. Is it going too far to say that both relationships are confirmed ... or am I missing something? 

Answer 1

While I do understand Kerry's https://www.wikitree.com/g2g/821271/why-should-the-mrca-pair-not-confirmed-when-you-triangulate?show=821441#a821441 answer and it's indeed correct from a statistical point of view, it's:

a) not 100% exact science, so there's a 1/64 % chance that you don't inherit DNA from both 2nd Great-Grandparents - so when it's not 100% I strongly advice against using it to confirm both 2nd Great-Grandparents.

b) even more important in my opinion is that we usually want to assign an ancestral DNA segment to a recent common ancestor so that we can then follow along that branch. When we're basically saying it can be both it means we don't know and that makes it not very helpful at all.

It's also going against the common practice that we don't confirm the MRCA couple (because again we don't know from whom we inherited) but stop at their 2 or more children from which we can follow the path downwards.

But that's just my two cents worth. 

Answer 2

Hi Ken,

See Kerry Larson's answer on this different but related question for a discussion of the reasoning for the WikiTree guidance for confirming to the MRCAs.

Comments

Rick and Ken,

I added a comment on Kerry's answer, in the hopes of sort of "de-bunking" some of those numbers that get thrown around. It's not really his fault - he's just repeating a misread of the one blog post that went around here, and making the mistake of assuming that certain "authoritative" sources are actually accurate. All that stuff is kind of besides the point, anyway (as I will get into with my own answer).

Answer 3

At the first cousin level, you can distinguish between a full 1C and a 1/2 1C based on the expected shared cM.  (Some might fall into a grey zone, though.)  It's not enough just to have the match, the match amount should be consistent with what is expected for a 1C and not a ½ 1C.  Based on that, you can confirm both MRCAs.  Same goes for second and third cousin relationships, though I think it really gets fuzzier for third cousins.

Answer 4

That's a good question to ponder, and "Yes, you're missing something".

In fact, this gets to the heart of why what you've described can only be done for 3rd cousins or closer. For more distant relations, it's commonplace for the shared DNA to only be from ONE of the two ancestors in the MRCA pair, and so in those cases (which require triangulation) it specifically states that you NOT confirm to the MCRA pair - only the path leading UP TO them.

Even with closer than 3Cs, there's a requirement that the match be of a "predicted 3C" level, which can have the effect of weeding out low 3C matches.

For 1st cousins, the shared DNA is typically from DOZENS of DNA segments. Each segment has a 50/50 chance of coming from a given grandparent. So if you have a typical number of segments - I'll use 3 dozen, or 36 - the chance of it all coming from just one of the grandparents would be 1/2 to the 36th power, which is 1 in 69 billion!

More than that, as Lisa points out, you'd only get half the cM in your DNA match. AncestryDNA does pretty well at identifying 1st cousin level matches, but a few get classified as 2nd cousin. One of my 1st cousins, for example, matches me 717cM, and is classified as "1st Cousin", but vs my brother she matches a bit lower, 606cM, and is classified as a "2nd Cousin". Their cutoff is apparently somewhere between 606cM and 653cM (one of my wife's 1Cs), so my 1C is just barely "over the line" into "2C" vs my brother.

The cutoff between 1C and below should probably be lower (I estimate 570cM would be good), which would make it so even few true 1Cs were misclassified (maybe only a few percent), but at the same time you would be a like number of matches from the next level being put misclassified into the 1C category. The two relationship levels overlap a little, so whatever line you draw will not be perfect.

So - for a true 1C - it's all good! You'll easily be declared a "estimated 1C", and you'll have plenty of segments to virtually guarantee that the shared DNA is from BOTH shared grandparents. Even for my 1C who was classified in the "2nd Cousin" category, the "Predicted Relationship" says "1st-2nd Cousin", so even SHE could be used for the confirmation.

But THAT's "where the plot thickens", doesn't it? (to be continued...)

Comments

"There's the rub!"

The top part of the "2nd Cousin" category on Ancestry DNA - down to somewhere around 350cM - says the "Predicted Relationship" is "1st-2nd Cousin". The highest 2C I've run into so far is 364cM, so matches in that range are generally 1C1Rs (1st Cousin, once removed).

In fact, the AVERAGE 1C1R is a bit over 400cM, so over HALF of the 1C1Rs will likely show up there.

The thing is, as far as cM numbers go, a half-1C and a 1C1R are the same thing - you share the same amount of DNA with both relations, on average. So over half of all half-1Cs would show up there, in the range above about 350cM that says "1st-2nd Cousin".

**** Which means that most biological half-1Cs could be used to do a DNA confirmation of BOTH your grandparents on that side, if the paper trail says that they're full-1Cs, even thought they're only descended from ONE **** In other words, DNA confirming an ancestor without any DNA from that ancestor. Not good.

So obviously, it's a system with a flaw (probably more than one flaw). In this case, what you could do to fix it would be to have a certain threshhold (for cM, for segments, or for both). That would make things more complicated I suppose (but in some ways it might be come LESS complicated, and just generally make more sense).

This sort of thing is talked about here periodically, so people KNOW it's not perfect. Small changes occur VERY infrequently. It's also been suggested that it be automated to the extent possible (which makes a LOT of sense).

The bottom line is that what we can do about it is (1) talk about it in places like this, hoping it will get around to somebody in a position to work on it, (2) demand our money back  and (3) most of all, do your own "DNA Confirmation" with some common sense. I've managed to find applicable DNA matches to confirm 7 out of 8 pairs of my gt-gt grandparents (for the last, there are no 3Cs to match!), but more importantly, the shared matches lists show all the people who are supposed to be on a given branch tend to match each other, and I have all kinds of matches that are 4Cs, 5Cs, and beyond corresponding to most of those gt-gt grandparents. My confidence in my biological lineage really isn't even based on the WikiTree rules! Don't solely rely on them (although they DO work reasonably well most of the time)!

Answer 5

Ken, of course you are correct.  The problem is the wording of the quoted advice.  Better wording would be, "If your match is around the mid expected range for a first cousin," say ~ 700 cM to 1000 cM, then it's highly or extremely likely that you match through both grandparents.  Half-first cousins, through one grandparent as in the situations you considered, almost never match this well.  At the higher end of that range, the odds are well over 99%.  At the lower end of the range, the odds are still very good.  We're talking confirmation here, of known family.

Are you at GEDmatch?  The best approach is to get to know your chromosomes through GEDmatch.  Recruit your first cousin to GEDmatch.  Then you can have certainty.  You will see where you match the first cousin at many chromosome locations through both grandparents.  Soon you'll be proving distant branches this way.