How to use Y700 to identify line of descent from known ancestor?

[gmatthews]

Practically everything I can find on the Big Y involves how to work with SNPs and sometimes how to use SNPs in combination with STRs to identify someone's ancestry. I also find a number of websites that say one can use the Y700 to determine which son of an ancestor someone descends from.

The latter is what I'm interested in doing. I keep up with about 35 or so individuals who mostly have tested at 37/67 markers who descend from a common ancestor who for me is 9 generations back and lived from ~1680 to 1762. He had 5 proven sons. Of those five sons, four are VERY well documented and there is little doubt (to me) that these four had any more sons that what I've already ID'd over the past 20 years. The 5th is very much undocumented although I believe one to three of his sons moved in a certain direction that the other families did not. I believe some of our group descend from this branch. I also postulate there was a 6th son of our MRCA from whom several individuals have tested to 67 or 111 markers. On top of that there are still a small number of others that I can't accurately place in our at-large tree.

So, there are several people in the 35 or so of us that we can't exactly fit into our extended family tree but based on a number of factors I'm certain that we all descend from the man who lived ~1680 to 1762.

Although all of us do have a "family marker" on one STR that no one else in our haplogroup has (last time I checked anyway) I still have not found any specific markers through 111 that can reliably say any specific person descends from any specific son of our MRCA.

To attempt to do this do I just use the Y700 for the STR markers and not deal with the SNPs? Or, as I (hopefully) get people to upgrade will our SNPs get more finely tuned? Right now there is only one other person in our group who has done the Big Y and the Block Tree shows our common SNP is 28 SNP generations back. We are much, much more closely related than that. Will getting more people tested reduce that number? I can't seem to find anything specific on how to try to ID lines of descent from a known ancestor or examples of people who have done something similar.

[benowicz]

The problem with STR markers is that there is always the potential for a "convergent" mutation masking the true time to most recent ancestor. Sometimes two donors coincidentally experience separate identical mutations on the same marker, or one donor experiences a subsequent mutation that effectively reverses an initial mutation. It can get even more complicated when you consider that supposedly something like one STR mutation out of 10 may be mutli-step--moving more than one allele value in a single event.

That almost never happens with SNPs. It is so incredibly rare for an SNP to mutate that for all practical intents and purposes you could say that "convergent" SNP mutations are impossible. I mean, you will find one or two in the literature, but the odds against it are mind boggling, whereas convergent STR mutations are routine for very old relationships.

I would almost always go with the SNP-derived conclusion, unless the genetic distances of STR profiles in the network are all very close.

Given the number of divergent SNPs typically involved, I would think it practically impossible that this 28 generation prediction via SNPs could actually be much closer than 25 or 23 generations. There is a lot more variation in the mutation rates of SNPs than there are for STR loci, but the span covered is also typically much longer. I can't imagine a case where the STR comparisons would yield a more accurate result than SNP comparisons.

Unless there is some incredible error here, like FTDNA totally screwing up the mutation rate for the relevant SNPs by a large exponential factor, or somehow reporting the results for a completely different donor, I would tend to conclude that there is just an unfortunate coincidence that both of you claim descent from the same man 9 generations ago or so. One of you seems almost certain to be wrong in that attribution, but by dumb luck just happens to belong to a remote branch of the same clade. I've seen that happen before, especially where the surname is common and where the region doesn't have a whole lot of haplogroup diversity, like Ireland.

[gmatthews]

Unless there is some incredible error here, like FTDNA totally screwing up the mutation rate for the relevant SNPs by a large exponential factor, or somehow reporting the results for a completely different donor, I would tend to conclude that there is just an unfortunate coincidence that both of you claim descent from the same man 9 generations ago or so. One of you seems almost certain to be wrong in that attribution, but by dumb luck just happens to belong to a remote branch of the same clade. I've seen that happen before, especially where the surname is common and where the region doesn't have a whole lot of haplogroup diversity, like Ireland.

There are roughly 20 of us who know how we connect back to our MRCA. There's another 7 or 8 who either trace back to the ancestor of the person that my Block Tree shows is 29 SNP generations back or who are related to me in some way that I don't know. I'm assuming I'm reading this tree correctly. Here's what I'm looking at:

2.jpg

I know there's no mistake in my paper trail and I know there's no mistake in the paper trail of the 20 other people I mentioned earlier who descend from our MRCA. Understanding that STR's can move up and down and backwards and forwards and that there's no way to tell when this happens or how often if you look at the standard STR panels of everyone in our group we're all pretty consistent in that there are certain markers that we all tend to have differences on and there are a ton that we all have identical results on. Some of those we differ on mutate fast and others slow. This guy shown in the above image is a better match to me out to 111 markers than some people that I'm only 5 generations from a common ancestor. He also matches almost identically to another STR tester who's paper trail I'm certain is correct and those two go back to the same common ancestor.

There's a lot of evidence that the ancestor of my Big Y match being very closely related to me in the way that I theorize which is that his ancestor is a son of a heretofore undocumented son of my MRCA and there is good evidence for that undocumented son (we know he exists but unlike the known sons there is no document that explicitly says he's a son of the MRCA).

So, all this aside, how do I use the Big Y to determine who descends from which son of our MRCA? There are other individuals who must descend from the MRCA that we've yet to be able to fit into our family tree definitively.

[benowicz]

Individual STR loci don't mean anything in isolation. I would just forget about that line of inquiry at this point. It may become useful in the future, but only within the special context of donors who can verify their common ancestry 9 generations back within the expected GD through SNP tests. There may be some interesting coincidences, but outside of the context of a donor whose relationship to you has been verified by an SNP-test, they mean nothing.

Quote: "This guy shown in the above image is a better match to me out to 111 markers than some people that I'm only 5 generations from a common ancestor. He also matches almost identically to another STR tester who's paper trail I'm certain is correct and those two go back to the same common ancestor."

I know it seems cruel, but I still trust the SNP based prediction way more than any STR based prediction. If his STR profile is way closer to you than many within the SNP-determined block of 20 who descend from the same common ancestor 9 generations ago, but this SNP-based test shows him related to you only 28 generations ago, I still think it is a matter of convergence and pure coincidence of surnames. I know that such a level of coincidence seems unthinkable, but it happens. Based on what I know now, my bottom line is that either that guy's STR GD to you is pure coincidence, and that he doesn't share a MRCA 9 generations ago, or that FTDNA royally screwed up reporting his SNP results.

As for this chart, I need more information to really understand it. I can't use it unless I know specifically which SNP block name the anomalous STR guy falls under and which SNP block is the most recent one shared by you and the other 20 donors. If you yourself really are block I-A13863, then it looks like the anomalous STR donor isn't pictured here, which is a problem because the picture wouldn't give me enough information to recalculate his TMRCA with you all via the SNP test. I need to know what is the most recent block shared by both the 20 donors and the anomalous STR donor, and how many positive SNPs he has below that block, unshared.

As for sorting out the specific lines of descent through the BigY SNP tests, it's pretty straightforward. The shape of the pedigree will approximate this SNP block sharing matrix: You are most closely related to the guy who shares the most SNP blocks with you. Hopefully your paper trails are detailed enough to sort through which son is represented by which SNP block. If there are any conflicts, then someone's paper trail is simply wrong.

I know this may sound discouraging, but never say 'never' as regards the potential for paper trail errors. I can't catalog every possible way that paper trails can get screwed up, especially with super common surnames, simply because they are endless. It's a bottomless pit that can only get straightened out through SNP testing.

[benowicz]

Just on the surface, if this anomalous STR donor really is under block I-A13863, then it looks like he has about the expected number of unshared SNPs for a TMRCA about 30 generations or so, plus or minus for variation in individual mutation rates. If he really is I-A13863, and FTDNA didn't just screw up reporting his SNP results, there is no way he could share an ancestor with you 9 generations ago. But there are two other I-A13863 donors who must share a common ancestor with him within the last 200 years. What do their STR results look like? Why aren't we talking about them? Are his SNP results correct but somehow his STR results incorrect? Again, this chart isn't clear to me. I need a clearer statement about which SNP block the anomalous STR donor falls under, and the most recent SNP block which is shared by all of the rest of you.

[gmatthews]

None of the other 20 have been Big Y tested. So far it's just me and the one other guy. That's why I was asking earlier if having more people in our group tested would help fine tune this. This is why I was asking in my first post do I need to focus on the STR's or SNPs to figure out which line those other people descend from, as I hopefully get each branch tested, so that we can tell how any anomalous tester fits into our tree. I don't want to start getting people to shell out $300 to $400 until I understand for myself how this works. They'll look to me to tell then what we need to do.

The other guy is also A13863 same as me. He's the 1 "DNA Match" shown in the box that says "your branch". Here is my haplo tree, what else do you need? The numbers to the left are the SNP generations back in time for that person's (surname in red to the right) haplogroup.

Matthews-BigY-Tree.jpg

[gmatthews]

But there are two other I-A13863 donors who must share a common ancestor with him within the last 200 years. What do their STR results look like? Why aren't we talking about them?

There are no others at A13863. Where do you see there are two more? If you're going by the number of countries shown, I don't know why it shows 3.

[benowicz]

Okay, much clearer. So you and this other guy are both I-A13863. One of you presumably is the British flag and the other the American flag. I have no idea what the question mark represents. Another tester whose results haven't fully posted yet? Let's just ignore the question mark guy for now.

This shows you sharing a common ancestor with him in maybe the last 200 years, +/- some standard deviation, given that you two average about 2 novel variants each. But that's about all you can tell from this. Precisely which son he descends from can't be determined from these results. Presumably it's a different son from the one you descend from, given the two novel variants.

I'm not sure what your earlier comment about a TMRCA 28 generations ago refers to now, especially if none of the other 20 have done BigY. Maybe you are referring to what the Y Full tree calls the "formed" date--the estimated age of the earliest SNP within the shared block. In this case, the approximate date I-A13863 diverged from I-A1465.

But that's not useful information. What you want is the TMRCA date, approximately the date of the most recent mutation within the shared block. You approximate that with the number of un-shared SNPs BELOW AND OUTSIDE OF the shared block. Which in the case of you and this other fellow, is 2 SNPs, which suggests that your most recent common ancestor was born about 200 years ago,+/-.

You could spend centuries farting around with analyses of individual STR loci, but without SNP tests, it would always be a guessing game. You could never really tell if some convergent STR mutations aren't distorting the picture.

I get your point about the relatively steep financial investment. A member of my family waited 10 years before finally breaking down and getting the BigY. But that's mostly because he doesn't have any statistically significant STR matches at all, and he's from a weird clade that was only discovered a couple years ago. With no other testers related more closely than 4,000 years ago, it was kind of pointless. In your case, the other testers would be almost guaranteed to get valuable information which could help them sort out which sons of the MRCA they descend from.

Sorry, but no pain, no gain. You could try some elaborate STR analysis to work through it, but with 300 years gone by since the MRCA, STRs would always leave a significant level of doubt about convergence.

[benowicz]

Anecdotally, I can say that I remember one fellow telling me that his STR profile registered as GD of 0 [Zero] at 67 against a fellow with whom he shared a MRCA born around 1550, versus a GD of 4 at 67 with a fellow with whom he had a MRCA born around 1750. I'm assuming that he was giving me an accurate account, which I believe is likely since the GD 0 fellow was supposedly born in Britain, and the GD 4 fellow an American, creating sort of a terminus ante quem. Don't know how typical this is, but it was enough to heighten the level of suspicion of STR results in my mind.

[gmatthews]

I'm not sure what your earlier comment about a TMRCA 28 generations ago refers to now, especially if none of the other 20 have done BigY. Maybe you are referring to what the Y Full tree calls the "formed" date--the estimated age of the earliest SNP within the shared block. In this case, the approximate date I-A13863 diverged from I-A1465.

But that's not useful information. What you want is the TMRCA date, approximately the date of the most recent mutation within the shared block. You approximate that with the number of un-shared SNPs BELOW AND OUTSIDE OF the shared block. Which in the case of you and this other fellow, is 2 SNPs, which suggests that your most recent common ancestor was born about 200 years ago,+/-.

The 28 came from me not understanding something Roberta Estes wrote in a May 30 blog post. Initially I thought the 28 generations was what you said: the age of the SNP, but I thought she was saying it was the genetic distance between me and the match which didn't make any sense but I went with it. Although I took the original Big Y test I never understood what it did and didn't put much effort into understanding it. I did the upgrade a couple of months ago and I'm now, just this week, trying to wrap my head around how all this works.

What do you mean when you say "You approximate that with the number of un-shared SNPs BELOW AND OUTSIDE OF the shared block." Do you literally mean the blue box below the box of 12 equivalent SNPS to A13863? The only "2 SNPs" I see is where it says that me and the Mathews match have an average of 2 private variants. Is that the way I would be able to put an estimate on the MRCA between me and another person? The average private variants and estimating 100 years per?

Thanks for your help, it is much appreciated!

[benowicz]

Yes. I just pulled the 100 years estimate from the sky, to make the math easy for myself. But I think that is supposed to be pretty close to the FTDNA average. It's not far from that average, at any rate. Given the standard deviation, it can't be far from it.