Is autosomal DNA testing sensitive enough to distinguish non-paternity events many generations ago?

Question

I have been researching my family line for decades, and have run into quite a brick wall with my earliest known progenitor, Daniel Cooper Sr. (1695-1795). There is an oft-repeated story about him being born at sea, but there is no evidence this actually happened. In addition, I'm a direct male descendant of Daniel and my Y-DNA doesn't match any Coopers besides known descendants of Daniel's, but my Y-DNA does match many folks in the Coryell family.

So if there was a non-paternity event, the question becomes, when? Daniel had 10 children, including 5 sons, but to the best of my knowledge all of the living male Cooper descendants come from Daniel Sr.'s first son, Daniel Jr. If I could find direct male descendants of one of Daniel Sr.'s other sons, I could do Y-DNA testing and we'd pretty quickly have some answers (or at least more data). But without those direct male descendants, can I do anything with autosomal DNA testing?

I'm thinking the answer is "No", but I would love to be proven wrong. Here's why I'm thinking autosomal DNA won't help. Imagine a family tree where woman B had one child with man A, and one whose father might be A, but might be C:

  A---B---C?A?
    |   |
D---E   F---G
  |       |
  H---I   J---K
    |       |
    M       P

So person E's DNA is 1/2 A and 1/2 B DNA (more or less, I know nothing is exact) and for now assuming person C exists, then person F's DNA is 1/2 C and 1/2 B. Continuing down, H's DNA is 1/4 A, 1/4 B, and 1/2 D, and M's DNA is 1/8 A, 1/8 B, 1/4 D, and 1/2 I. Similarly, P's DNA is 1/8 C, 1/8 B, 1/4 G, and 1/2 K.

But if there is no C - if F's father is A - then P's DNA is 1/8 A, 1/8 B, 1/4 G, and 1/2 K.

That means that at 3 generations down, M and P share 1/4 of their DNA if A if the father of F, but only 1/8 of their DNA if C is in fact the father of F. And in this case, Daniel Cooper Sr. is my great(5)-grandfather, so if I find a non-male descendant of one of his other children, that person and I have either 1/64 or 1/128 matching DNA and that just doesn't seem like enough to show up on an autosomal DNA test.

Or am I thinking about this the wrong way?

Answer 1

In order to do what you're trying to do with autosomal DNA, with an ancestor that far back, you would need to (a) have an extremely large set of matches to work with, across all the descendant lines, and (b) probably be able to trace specific significant-sized segments down the family tree using chromosome browser tools.

The likelihood of having any match whatsoever at the 4th cousin level is about 50%; for 5th cousins, that drops to abt. 12.5%, and goes down by another factor of 4 for each subsequent generation (the ISOGG "averages" for these relationships are necessarily skewed on the high side, because it's difficult to fully account for all the non-matches for each distant cousin relationship). As the relationships get more distant, it would take a correspondingly larger number of samples to generate enough significant matches.

Cross-branch matches don't really help, unless all of the pedigrees have been traced far enough, and accurately enough, that you can rule out any other common ancestors for any of the pair-wise matches. That can be especially difficult if the family lived in the same place for multiple generations, increasing the likelihood of intermarriage with other ancestral families.

You would need to establish triangulation of significant-sized segments (above 15 cM, at least, and preferably above 20 cM) and then trace those segments backward and forward through the family tree. Bear in mind that a 10 cM segment has a 50% chance of being IBC (inherited by chance), a 7 cM segment about 80% and a 15 cM segment abt. 20%.

I have a similar situation with presumed descendants of an ancestor b.abt.1765. There are very few matches across descendant lines large enough to establish that all (or any) of the hypothetical children are in fact siblings, and the matches we do have still aren't large enough for tools like WATO to be useful. About the best the statistical calculations suggest is that for any given hypothetical child line, if all the other child lines are correct, then the target line is, more likely than not, also correct. But that's not a particularly compelling result (it's also possible that none of them are correct).

You might get slightly better results with X-DNA, as long as you have lines that don't have two generations of males in succession that you can work with. X-DNA is more likely to get passed without (or with less) recombination than a-DNA. But you;d still need a chromosome browser to track segments.

For ancestors as far back as you're dealing with, y-DNA is a much better bet, assuming you have multiple patrilineal lines to work with.

Answer 2

I'm not sure that simple percentages are that useful but The Shared cM Project lists the centimorgans of common autosomal DNA that can be expected for various relationships. A common 5th great grandparent will mean you are 6th cousins. The chart on that page shows that the "average" common cM between 6th cousins is 18. As I understand it, these are numbers that have been found in real life. However, the range for that common cM goes from 0 to 71. Again, this is the range that's found in real life though it's a range that applies in 90-something percent of real life.

So the point is that if you have zero cM in common with someone who, on paper, is your 6th cousin, that doesn't mean that you aren't 6th cousins, because zero common cM is known. (Of course, there may be some circular logic involved here)

Further, the blunt truth is that with a limited number of letters and phrases in the DNA alphabet, DNA recombination may result in unrelated people having common sequences by chance. Ancestry won't display common cM lower than 8 (unless it's 6, sorry) as people (not Ancestry) reckon that a match at that lower limit is 50/50 likely to be Identical by Chance, rather than Identical by Descent.

So, if you have a supposed 6th cousin (on paper) then it's possible that you have no reportable common cM of DNA but you are actually genuinely cousins. It's also possible that you are reported as having 8 cM (or whatever it is) suggesting that you might be cousins but this is incorrect and you are not related.

I have only a basic understanding of DNA but that's my two pennorth, as it were. All comments gratefully accepted.