I have a DNA match to someone on Geneanet with 12.5 cM shared. I happen to know him (he is in my genealogy society) so I asked him for his GEDmatch kit number. Both of us uploaded to GEDmatch with Family Tree DNA raw data and the same to Geneanet. However, GEDmatch shows no matching segments at all (even with the default minimum segment size of 7 cM). How could Geneanet's matching algorithm be so far off GEDmatch's?
Don Worth GEDmatch kit # = T589519
Phil Crow GEDmatch kit # = T296090
DNA match oddity/problem
-
- Beiträge: 2
- Eingabeform: Grafisch
- Navigation: Grafisch
- Den Stammbaum ansehen
Different companies use different algorithms with different parameters. It's difficult to identify endpoints, and becomes an arbitrary decision, that can cause differences in segment lengths, as well as defining how much change is allowed. One company may allow only one mutation/change within a section, where another company may allow more, and this would make a big difference in whether a single segment is identified, or it's separated into 2 segments at the change, or the first segment is just ended/shortened at that point.
I prefer to view comparisons at much smaller segment lengths, fully knowing that most very small segments are likely random chance matches, but there's a little more to learn anyway. Here's yours at a minimum 3cM length:
Comparing Kit T296090 (Phillip Crow) [Migration - F2 - T] and Kit T589519 (Don Dyson Worth)
Chr B37 Start Pos'n B37 End Pos'n Centimorgans (cM) SNPs
2 129,983,541 132,428,421 3.3 357
4 185,520,068 186,372,826 3.1 219
5 4,376,717 5,209,872 3 284
6 99,203,222 103,325,998 3.1 746
7 8,161,981 9,214,578 3 456
7 29,450,895 31,003,819 3.2 474
7 137,669,251 141,669,428 6.9 910
8 101,691,690 103,048,115 3.2 368
12 54,534,721 56,829,092 3 470
16 49,962,406 51,451,319 3.1 341
17 9,026,077 9,667,597 3.1 201
17 30,258,656 31,967,460 3.6 425
19 1,876,566 2,950,607 5 293
19 34,315,896 35,655,151 3.1 295
22 37,081,695 37,671,677 3.1 229
Largest segment = 6.9 cM
Total Half-Match segments (HIR) 52.6cM (1.466 Pct)
15 shared segments found for this comparison.
649233 SNPs used for this comparison.
You just missed the 7cM minimum with that 6.9 on 7. You also have a 5 on 19, and a lot of other small segments. There are no almost contiguous segments, so not an arbitrary algorithm split. (I was looking for something like a 7 followed immediately by a 5.5.)
Only Geneanet can say why they identified a 12.5cM segment, and I suspect their developer will find your comparison interesting, may help them tweak their algorithm.
In my view, when I find cases like this with so many small segments, it's highly likely that you are related but very distantly. You probably share a distant common population where both your ancestors lived and married for a long time. That won't help you find matches in recent times, within the last 6 or so generations, but there's probably considerable shared ethnicity.
[I suspect someone may want to move this to the DNA board.]
I prefer to view comparisons at much smaller segment lengths, fully knowing that most very small segments are likely random chance matches, but there's a little more to learn anyway. Here's yours at a minimum 3cM length:
Comparing Kit T296090 (Phillip Crow) [Migration - F2 - T] and Kit T589519 (Don Dyson Worth)
Chr B37 Start Pos'n B37 End Pos'n Centimorgans (cM) SNPs
2 129,983,541 132,428,421 3.3 357
4 185,520,068 186,372,826 3.1 219
5 4,376,717 5,209,872 3 284
6 99,203,222 103,325,998 3.1 746
7 8,161,981 9,214,578 3 456
7 29,450,895 31,003,819 3.2 474
7 137,669,251 141,669,428 6.9 910
8 101,691,690 103,048,115 3.2 368
12 54,534,721 56,829,092 3 470
16 49,962,406 51,451,319 3.1 341
17 9,026,077 9,667,597 3.1 201
17 30,258,656 31,967,460 3.6 425
19 1,876,566 2,950,607 5 293
19 34,315,896 35,655,151 3.1 295
22 37,081,695 37,671,677 3.1 229
Largest segment = 6.9 cM
Total Half-Match segments (HIR) 52.6cM (1.466 Pct)
15 shared segments found for this comparison.
649233 SNPs used for this comparison.
You just missed the 7cM minimum with that 6.9 on 7. You also have a 5 on 19, and a lot of other small segments. There are no almost contiguous segments, so not an arbitrary algorithm split. (I was looking for something like a 7 followed immediately by a 5.5.)
Only Geneanet can say why they identified a 12.5cM segment, and I suspect their developer will find your comparison interesting, may help them tweak their algorithm.
In my view, when I find cases like this with so many small segments, it's highly likely that you are related but very distantly. You probably share a distant common population where both your ancestors lived and married for a long time. That won't help you find matches in recent times, within the last 6 or so generations, but there's probably considerable shared ethnicity.
[I suspect someone may want to move this to the DNA board.]
My page: https://www.wikitree.com/wiki/Jacobson-1250
My DNA: https://www.wikitree.com/wiki/Jacobson-1250#DNA_Evidence
yDNA Haplogroup: R-S23346 https://www.wikitree.com/wiki/Space:YDNA_Haplogroup_R-S23346
mtDNA Haplogroup: K1c2 https://www.wikitree.com/wiki/Space:K1c2_mtDNA_Haplogroup
mtDNA Help: https://www.wikitree.com/wiki/Space:K1c2_mtDNA_Haplogroup#About_Your_mtDNA_Test_and_Haplogroup
My DNA: https://www.wikitree.com/wiki/Jacobson-1250#DNA_Evidence
yDNA Haplogroup: R-S23346 https://www.wikitree.com/wiki/Space:YDNA_Haplogroup_R-S23346
mtDNA Haplogroup: K1c2 https://www.wikitree.com/wiki/Space:K1c2_mtDNA_Haplogroup
mtDNA Help: https://www.wikitree.com/wiki/Space:K1c2_mtDNA_Haplogroup#About_Your_mtDNA_Test_and_Haplogroup
-
- Beiträge: 2
- Eingabeform: Grafisch
- Navigation: Grafisch
- Den Stammbaum ansehen
Thanks for spending time on this. I appreciate it.
I've seen what Ancestry's Timber algorithm can do to segments but for some reason it didn't occur to me to look on GEDmatch at segments below 7 cM. I guess it was because I'm pretty sure GEDmatch doesn't take out pileup regions so the explanation would have to have more to do with fuzzy ends of segments. It just seemed to me that going from 12.5 cM to sub-7 cM is a pretty drastic difference - especially since you found that the smaller segments weren't adjoining one another and we are not talking about a comparison between different sequencer chips.
I suppose I should look at some of my other matches on Geneanet to see if they, too, vary a lot from GEDmatch. But most of them are rather distant (sub-20 cM) and it may be hard to correlate them to matches I have on GEDmatch (or any of the other 5 companies I have tested at).
Thanks again!
Don
I've seen what Ancestry's Timber algorithm can do to segments but for some reason it didn't occur to me to look on GEDmatch at segments below 7 cM. I guess it was because I'm pretty sure GEDmatch doesn't take out pileup regions so the explanation would have to have more to do with fuzzy ends of segments. It just seemed to me that going from 12.5 cM to sub-7 cM is a pretty drastic difference - especially since you found that the smaller segments weren't adjoining one another and we are not talking about a comparison between different sequencer chips.
I suppose I should look at some of my other matches on Geneanet to see if they, too, vary a lot from GEDmatch. But most of them are rather distant (sub-20 cM) and it may be hard to correlate them to matches I have on GEDmatch (or any of the other 5 companies I have tested at).
Thanks again!
Don