Why is GEDmatch report omitting some matches?

6

I ran a GEDmatch multi-kit analysis and got this report for Matches with Joseph's kit, for chromosome 12:

enter image description here enter image description here

And putting a different kit in first position, I get this report for matches with Charlotte:

enter image description here enter image description here

The strange thing I notice is on Chromosome 12, in the first report I get several matches that span as much as from 1,359,590 to 5,316,226 (about 12.3 cM). In the second report, I get Edna matching Charlotte from 1,517,192 to 3,941,530 (7.58692 cM).

The segment matches Charlotte from the first report, so why doesn't Edna appear in the first report? Is this a bug?


NOTE: All kit numbers have been removed, and names changed to pseudonyms for ease of discussion.

WilliamKF

Posted 2017-04-10T22:43:19.623

Reputation: 5 145

1

Would it be possible to obscure the names associated with these kits? See this post from The Legal Genealogist: http://www.legalgenealogist.com/2015/01/25/9381/ "...we should not take a screen capture of DNA results from a testing company and post it in a blog post or on Facebook with the names or pictures of our matches still attached unless we’ve asked those matches specifically if we can post it." Linking to results isn't the same as a screenshot, but I think the same caution should apply.

– Jan Murphy – 2017-04-10T23:25:43.870

William – I've edited the question to remove the names and kit numbers, and are happy to reopen it now there is no identifying information. I've replaced all the names with pseudonyms. – Harry Vervet – 2017-04-11T12:51:06.950

@ColeValleyGirl I think this warrants further discussion on Meta as to what our policy on this is. Under the current revision someone could still argue that it contains DNA data without their permission. Providing a useful answer may now be difficult because it doesn't contain enough detail. – Harry Vervet – 2017-04-11T13:45:14.620

@WilliamKF What cM threshold did you use for each analysis? Might a simple explanation might be that the match between Edna and Joseph is below the threshhold you used? – None – 2017-04-11T13:54:13.527

@ColeValleyGirl Yes, we need the discussion, because there are a lot of DNA questions that need heavy editing and redaction if this is going to be the policy going forward. It is still not clear whether the amount of data currently included in the questions is "breaking the terms of service of other sites". – Harry Vervet – 2017-04-11T13:55:02.103

2So little information remains in the edited question that I doubt much can be said about the particular situation. – RobertShaw – 2017-04-11T19:46:41.183

@ColeValleyGirl The threshold was the default multi-kit analysis setting of 5 cM, so that doesn't explain the omission. – WilliamKF – 2017-04-11T23:52:30.397

See Meta discussion here.

– WilliamKF – 2017-04-12T00:04:07.013

@WilliamKF if you examine the matches for each kit using the one-to-many tool, do you get a similar anomaly? – None – 2017-04-12T09:12:13.227

Is http://blog.kittycooper.com/2014/09/using-gedmatch-with-an-ancestry-kit-the-gedmatch-chromosome-browser/ helpful, including the statement that overlapping segments do not always mean a match?

– None – 2017-04-12T09:32:58.223

@ColeValleyGirl I reviewed Kitty Cooper's blog entry and to my understanding the overlapping segments not always matching means you can get false positives, not that matches are not transitive. – WilliamKF – 2017-04-13T18:42:04.290

@WilliamKF Lkessley as pointed you to more resources that explain the situtation. Have you had a chance to look at the one-to-many matches yet? – None – 2017-04-14T07:37:07.033

Directly from GEDmatch email in response to a query: In terms of copyright infringement we are not concerned about showing data from GEDmatch. It would be nice if we are identified contd. – None – 2017-04-14T09:09:30.117

However privacy is a different matter. Many people are concerned about having their DNA identified, especially linked to their name in any way, including a kit number that might be traced. Several blogs use GEDmatch data and screen shots but never with identifying information. I strongly recommend you get permission if possible. Bottom line: protect from identifying others unless you have their permission and act in such a way that GEDmatch doesn't get any complaints. – None – 2017-04-14T09:09:36.883

On that basis, I'm giving @WilliamKF the benefit fo the doubt and assuming he asked for permsission and reverting my edit. – None – 2017-04-14T09:14:06.777

Answers

5

There are 46 chromosomes that come in 23 pairs, one from mother and one from father. You are talking here about chromosome pair 12.

The way matching works is that it cannot determine whether the match is occurring from the maternal chromosome or from the paternal chromosome. It only determines that either the maternal or paternal chromosome is matching at each point along the chromosome pair.

In your case, on that segment on chromosome pair 12 you have:

  • Joseph matches Laura
  • Joseph matches Charlotte
  • Charlotte matches Joseph (good, that verifies)
  • Charlotte matches Laura
  • Charlotte matches Edna (7.6 cM)

There are two possible reasons why Edna does not match Joseph.

  1. Charlotte matches Joseph on one of her chromosomes (maternal or paternal), and Charlotte matches Edna on the other.

  2. Charlotte and Edna match not because it is a common segment, but because it is a by-chance match where matching base pairs crisscross between the paternal and maternal chromosomes and appear to the matching algorithm to line up. The smaller the matching segment, the more likely this is to happen. Normally you need a 15 cM segment or longer before by-chance matches become improbable. At 7.6 cM, you always need to be suspicious that this might be a by-chance match.

Laura matches both Joseph and Charlotte so they triangulate meaning the 3 of them match each other on the same segment. Triangulation (without a by-chance match in there) is good assurance that they are matching on the same chromosome.

To verify whether Edna is matching on the other chromosome, or whether Edna's match with Charlotte is a by-chance match, you need other people to compare to. You will have to find at least one other person who matches Charlotte on that segment but does not match Joseph and Laura. Then on GEDmatch, you can do a one-to-one compare of that new person with Edna. If the new person matches Edna, then Charlotte, Edna and the new person all match on the same segment and triangulate. They will be matching on Charlotte's other parent's chromosome from the one she is matching with Joseph and Laura.

If the new person does not match Edna, then either Edna, or that new person, or both are a by-chance match with Charlotte and you'll need to find yet another person to compare with.

The key is to realize (1) there are two chromosomes in each pair that a person can match to, and (2) by-chance matches on small segments are possible. Knowing this allows you to logic through the possibilities.

lkessler

Posted 2017-04-10T22:43:19.623

Reputation: 16 148

It was my first thought too that they matched on opposite chromosomes, but looking at the data (which has since been edited out of the question) I came to conclusion this was unlikely – Harry Vervet – 2017-04-12T14:54:23.317

1@HarryVervet interested to know how you could tell that from the 2D chromosome browser. – None – 2017-04-12T17:35:51.137

@ColeValleyGirl I would have to look at the data again but it was a combination of the overlap of several of the matches and length of segments that made me think this. Coincidences happen, though, and I could be wrong. – Harry Vervet – 2017-04-12T18:22:00.027

This answer contradicts my current understanding of DNA, can you provide a source to cite your argument that in any given segment range, there are two sides to the segment and a match could be on one side or the other, and thus segment matching is not transitive? (Note: I do agree that for any given segment, it may consist of portions from your dad and other portions from your mom.) – WilliamKF – 2017-04-13T18:47:15.743

1

@WilliamKF - Kitty Cooper describes it quite well here: http://blog.kittycooper.com/2014/10/when-is-a-dna-segment-match-a-real-match-ibd-or-ibs-or-ibc/ She gives an example of Dad giving her AAAAAAAAAAA and Mom giving her CCCCCCCCCCC and how she would appear to match to anyone with either A's or C's on that segment, e.g. ACACCAACCAC or CCAACCCACA. So using that example, people with AAAAAAAAAAA on that segment will match to each other on her dad's side, and people with CCCCCCCCCCC on that segment will match to each other on her mom's side.

– lkessler – 2017-04-14T04:44:47.937

2

@WilliamKF - Roberta Estes gives a similar example here: https://dna-explained.com/category/ibc-identical-by-chance/ and talks about an identical by chance match due to "zigzagging back and forth between your Mom's and Dad's DNA strands."

– lkessler – 2017-04-14T04:53:28.950

1@lkessler Thanks, didn't realize the technology was doing an OR and thus allowing a match to hop back and forth between your mom and dad. Hopefully they will improve the sequencing in the future to map the two sides of a chromosome independently, which would make a match transitive. – WilliamKF – 2017-04-15T02:38:18.380