MyHeritage made a big splash at RootsTech last week highlighting its new chromosome browser which initial reports indicate is nicely constructed, robust, and includes a true triangulation feature. The debut of the browser was in conjunction with the announcement of major improvements in the MyHeritage matching procedures and algorithms.
The announcements came not at RootsTech—which wrapped-up last Saturday in Salt Lake City—but a few weeks earlier on the MyHeritage Blog. It was at the massive RootsTech though, this year estimated to have had over 14,200 paid attendees, where attention on MyHeritage became front and center.
Anyone who took a MyHeritage DNA test, and anyone who uploaded DNA data from another service, will now receive more accurate DNA Matches; more plentiful matches (about 10x more); fewer false positives; more specific and more accurate relationship estimates; and indications on lower confidence DNA Matches to help focus research efforts.
—MyHeritage Blog, 11 January 2018
Mind you this is informal, preliminary, and nothing but a cursory glimpse, but MyHeritage's improvements had some well-known names take a quick look.
Debbie Kennett, on her blog yesterday morning, posted an item about the MyHeritage updates. Debbie's review of the new chromosome browser was favorable, but she noted two areas overall at MyHeritage that might possibly need more investigation.
The first deals with genotypical pile-up regions. One area noted was near the start of chromosome 15 (not a region currently known to be one of broad-population excess IBD sharing), and there was associated discussion about it here on the Genetic Genealogy Tips & Techniques Facebook group. Debbie found another pile-up region on chromosome 3 with 15 people shown as triangulating.
She advised caution when drawing conclusions from the triangulated segments as viewed in the browser because it's "not just the seven matches you can see in the chromosome browser that you need to consider but also how many other people share the segment. The more frequent the segment is in the population the less likely it is to fall within a meaningful genealogical timeframe." Sage advice.
The second area of possible concern was echoed by Blaine Bettinger yesterday morning in these posts to his Genetic Genealogy Tips & Techniques Facebook group. The quick summary is that Blaine reports now showing 3,655 matches at MyHeritage; 2,729 are shared with one or both parents, and 926 are shared with neither parent. That's 25.3% of the total matches. Blaine's kit used for this was a MyHeritage test; both his parents' data were uploaded from AncestryDNA tests. He noted that the source of the kits used for the comparison could affect the results significantly, and that he plans to follow-up by using his own AncecstryDNA results so that all three kits would be from the same company performed on the same chipset.
But the possibility of one-in-four false matches is enough, at least preliminarily, to raise an eyebrow. Debbie didn't specify the testing-company/version for the source of her data files at MyHeritage, but she reports very similar results as Blaine. Of her 2,063 matches, 768 match her father and 699 match her mother. That means 28.9% of her reported matches matched neither her mother nor her father, and could possibly be false.
Only some genealogists have the benefit of being able to use autosomal test data for both parents, but those who do can provide the rest of us with something of a baseline. I don't think anyone is drawing any conclusions yet. We'll need more information. But if the 25% to 29% range holds true as likely false due to matches reported to the child but to neither parent, there could be an issue for MyHeritage to address.