Explaining major discrepancy in genomic position between Gedmatch and MyHeritage?

Question

I have identified a discrepancy with several test kits that were tested at Ancestry, with results transferred to both Gedmatch and MyHeritage, all relating to the same area at the end of chromosome 19.

For one segment match on chromosome 19, Gedmatch and MyHeritage show the match at two different close but non-overlapping locations:

Gedmatch:

MyHeritage:

As you can see, Gedmatch states the match is in the 59M-63M position, while MyHeritage shows this segment in the 54M-59M position. When going to map these results in DNA Painter, they therefore appear to be two distinct segments, when in reality they are actually the same matching segment (the blue and green should be overlapping):

The other segment matches between these two people all are very similar in terms of genomic position, on both Gedmatch and MyHeritage.

Now small differences I could live with, since each service uses different SNPs. But there is something not right here.

How to explain the discrepancy? Is this occuring due an error on the part of Gedmatch, MyHeritage, or with the actual kit data from Ancestry?

Answer 1

It looks like this may be due to a difference in the Genome Builds (https://en.wikipedia.org/wiki/Reference_genome#Human_reference_genome) used between the two tools.

GEDMatch reports that they store data in Build 36 format (https://www.gedmatch.com/gedwiki/index.php?title=DNA_Upload). Louis Kessler noted that MyHeritage raw data is downloaded in Build 37 format (http://www.beholdgenealogy.com/blog/?p=2136), suggesting that the MyHeritage analysis would also be in that build.

For the SNPs shown in your MyHeritage image, rs8112586 has a Build 36 position of chr19:59,276,190, but a Build 37 position of chr19:54,584,378. (Positions determined using Human Assembly Build 36 or Build 37 for searches at UCSC Genome Browser.) The second SNP shown in MyHeritage (rs117625253) does not show up in Build 36 of the genome browser, but the Build 37 position for it is chr19:59,097,933 (and this position corresponds to chr19:63,789,745 in Build 36 using UCSC's LiftOver tool).

So, it seems like it actually is the same segment, just appearing differently depending on which Genome Build you view it in - Build 36 says chr 19 59-63M, and Build 37 says chr 19 54-59M.