PDA

View Full Version : One-to-one matches cM vs #SNPs



therrien.joel
09-10-2017, 12:41 AM
So I'm curious. When someone does one-to-one matching, at least on GEDmatch, you get the matches listing how long the segment is in cM and how many SNPs there were in that segment. I get the idea that those two are roughly related but that it varies depending on the chromosome and I guess even along a chromosome. So when you are looking at more distant matches where these numbers get smaller, is there any rule of thumb for which number is a better indicator that a match is IBD as opposed to IBC?

Saetro
09-11-2017, 11:41 PM
Basically, if they are both small, the match is highly unreliable.

Curious is good, but can be frustrating, time wasting and just plain inefficient when applied to small matches.
Leave them alone!

Work on the big end - above 20cM first and then above 15cM. (This is pain and disappointment talking.)

The problem is that many of your 10cM matches are 10 generations ago.
Now sometimes that is great, because other evidence points to places I can look for suitable documentary data on links.
And sometimes they are in the middle of segments that I have already linked to one branch of family (and they are not on the other pair of the chromosome.)

But odd small fragments are just a waste of time.
For a start, around 5cM 80% are false matches.
See ISOGG https://isogg.org/wiki/Identical_by_descent
And you don't know which ones! (Described as a bowl containing 80% poisoned M&Ms.)
This reference shows that if you want to use small segments, then DNA test your parents and then phase your results.

But to answer your question more directly.
If you are looking at segments that are smaller than about 7cM and 500SNPs then they are more likely to be false matches.

therrien.joel
09-12-2017, 02:59 AM
Basically, if they are both small, the match is highly unreliable.

Curious is good, but can be frustrating, time wasting and just plain inefficient when applied to small matches.
Leave them alone!

Work on the big end - above 20cM first and then above 15cM. (This is pain and disappointment talking.)

The problem is that many of your 10cM matches are 10 generations ago.
Now sometimes that is great, because other evidence points to places I can look for suitable documentary data on links.
And sometimes they are in the middle of segments that I have already linked to one branch of family (and they are not on the other pair of the chromosome.)

But odd small fragments are just a waste of time.
For a start, around 5cM 80% are false matches.
See ISOGG https://isogg.org/wiki/Identical_by_descent
And you don't know which ones! (Described as a bowl containing 80% poisoned M&Ms.)
This reference shows that if you want to use small segments, then DNA test your parents and then phase your results.

But to answer your question more directly.
If you are looking at segments that are smaller than about 7cM and 500SNPs then they are more likely to be false matches.

Thanks, this is good info. I'll admit I was trying to see if I could do something that was closer to identical by population... i.e. see if one population has more hits than another, but I suspect you need substantial statistics to have that work or it just plain won't.

So then a dumb question, if you don't mind: I accidentally ran across a person that has both his parents and himself listed. Obviously he and his dad match each other. But they both match me on the same chromosome and same location but only at 3.3cM. So, still very probably just garbage, right?

Saetro
09-13-2017, 12:02 AM
So then a dumb question, if you don't mind: I accidentally ran across a person that has both his parents and himself listed. Obviously he and his dad match each other. But they both match me on the same chromosome and same location but only at 3.3cM. So, still very probably just garbage, right?

Not dumb at all.

I have seen a guru comment on this not too long ago.
(But of course I can't find the comment right this minute.)
I seem to recall that he said that this is still very likely to be false.
(In the example the questioner was the person who had tested his parents.)
What he said was needed was someone like a second cousin to that person, who also matches on that segment.

firemonkey
09-13-2017, 01:35 AM
I've contacted several people with shared segments 15cMs+ . I've yet to make any connections.

therrien.joel
09-13-2017, 01:42 AM
Not dumb at all.

I have seen a guru comment on this not too long ago.
(But of course I can't find the comment right this minute.)
I seem to recall that he said that this is still very likely to be false.
(In the example the questioner was the person who had tested his parents.)
What he said was needed was someone like a second cousin to that person, who also matches on that segment.

That makes a lot of sense. Of course the chance of finding that set of people goes way down.

I will say that the link you gave has really been very helpful on making more sense of when one can try using small segments. In particular the links to DNA explained at the bottom of the page that discuss this:
https://dna-explained.com/2015/01/11/just-one-cousin/
https://dna-explained.com/2015/01/21/a-study-utilizing-small-segment-matching/

About that family I found on GEDmatch. I dug deeper into it just for curiosity's sake. Harmless so long as you don't loose your bearings. Inspired by the DNA explained articles, I thought well why not check for small segments? I ended up finding six segments that "triangulate" between me and either the brother or the sister (in one case both) and either the father or the mother. Now that last part sounded like it was the proof that this was nothing other than a coincidence until I checked the mother against the father and found they are something like second-cousins. So obviously I am not by any stretch saying this family is related to me, but it is interesting. Again, just to be clear, the prime reason I am looking at small segments is because any link is going to be distant enough that larger segments are simply not going to exist, so I have little other choice and am maintaining an awareness of exactly what limitations this places on what I can discover.

Edit: OK dummy me... there were two people on the GEDmatch's one to many page for the brother that were at the ~2'nd cousin level. For one of the segments that I match with him, he also matches with both of them and his mother.

therrien.joel
09-13-2017, 02:15 AM
I've contacted several people with shared segments 15cMs+ . I've yet to make any connections.

I'm certainly not holding out any hope!