PDA

View Full Version : 23andme Relative Finder and FTDNA Family Finder Thresholds



J Man
04-14-2013, 03:22 PM
I know that this is still not an ''exact science'' but what is the lowest threshold in cM size for determining if a Relative Finder or Family Finder match is truly IBD ie. real or not?

geebee
04-14-2013, 05:53 PM
This is just an opinion -- I suppose a guess, really -- but I think some small segments are likely IBD while others of the same size are not. So I don't know that you can give a number, unless you're asking something along the lines of "how many cM does a segment have to be to be SURE it's IBD, and not IBS".

But, as far as IBS segments themselves are concerned, my understanding is that they are more likely to appear when two people have the same ethnic background. If this is correct, I don't think it's likely to be coincidence. My guess would be that even though a particular segment thought to be IBD may turn out to be IBS, being made up of smaller segments from both chromosomes, those smaller segments may actually be IBD.

What do I mean? Well, in the first place, everyone is related to everyone else in some degree or another. There really is no such thing as two "unrelated" people. But, in the second place, people from the same ethnic group are likely to be more closely related -- all other things being equal -- than people of different ethnic groups.

So these false IBD segments actually are an indicator, of sorts, of a greater "background" relatedness. In addition, I would argue that people never have just one set of "common ancestors" (well, of course they also have all the ancestors previous to those, but that isn't what I mean). Rather, because there aren't enough people in the world to provide all of our "theoretical" ancestors, we're all related to each other in many, many ways.

It's my guess that the little segments are actually signals of that, in many cases. So if, say, I show a "pseudo segment" with someone, composed of smaller segments from both parents, it may not be entirely random. Yes, for any given SNP there may typically be only two possibilities, so very small matching segments relying on unphased chromosomes may well be mere chance.

But suppose a 7 cM segment is actually composed of several 1 cM segments, plus "filler" (a few SNPs between the segments that really are just random matches). Even 1 cM segments may not match simply by chance. It's just that they may represent connections so old that they aren't genealogically meaningful. But they aren't entirely IBS, either. They truly reflect common ancestry.

Of course, this is just my speculation, and I realize it doesn't actually answer your question. If someone can, I would certainly be interested in the answer.

J Man
04-14-2013, 09:42 PM
^I do see what you are saying here geebee. Even very small segments down to 1 cM may truly be IBD and reflect ancient shared relationships. I am wondering about the larger one's though.

geebee
04-15-2013, 03:10 AM
For whatever this may be worth, I think 23andMe uses 7 cM and 700 SNPs to determine what to report in Relative Finder -- although this changes once you're actually sharing with someone. So this may be what they see as the "theshold".

If you're already sharing with someone, they will report smaller segments -- even down to 4 cM. I guess their thinking is that if you already have reason to believe you might be related to someone, smaller segments may be more meaningful.

Of course, don't forget that they're using unphased data. For unphased data, you could rely on even smaller segments. But, I would say that at 7 cM a segment is more likely to be IBD than IBS, and as segments went up in size, they would be even more likely to be IBD. Down in size, they might still be IBD, but become less and less certain.

But I'm sure someone will eventually weigh in who has a better understanding of this than I do. :)

EDIT: Since I have data for a few folks in three different generations of one family, I consider it lucky when someone matches both my father and me or my daughter and me. Or even if the person matches either or both of my siblings and me -- if it's in an area where we only have half matching.

These combinations make it more likely that the segment is a real match -- using just one chromosome, not smaller segments from both. When that happens, I feel pretty confident about any reported segment (even 4 cM).

However, I've also occasionally found segments of 7 cM or even slightly larger that match my daughter but neither parent. This could be due to the uncertainty regarding start and end points -- so that it perhaps does match one of the parents, but isn't showing because it's just under the threshold.

The problem with that is the fact that even a smaller segment -- down to 4 cM at least -- should still be showing in Family Inheritance, even if that would not by itself be enough to include it in Relative Finder.

So if my daughter's segment says 7 cM or more but neither parent has even a 4 cM segment, I tend to discount it. It might be composed of tinier IBD segments -- in fact, that may be likely -- but I agree with most folks who think that while these may reflect relatedness, it may be beyond any possible genealogical time frame.

(After all, if the documents existed and could be found, in theory it should be possible to construct a family tree linking any two people in the world, righ?)