Ancestry tests are becoming increasingly popular with more and more people eager to discover their heritage. But the science can be somewhat dubious.
Even Paul Maier, the chief geneticist at FamilyTreeDNA, suggests we should all take the results with a generous pinch of salt by admitting it is "kind of a science and an art". And the experience of Charlsie Agro, host of the Canadian show Marketplace, and her identical twin goes to show exactly that.
Charlsie and her sister, Carly, bought home DNA kits from five different brands: AncestryDNA, MyHeritage, 23andMe, FamilyTreeDNA, and Living DNA. Next, they took samples of their DNA to send back to the companies for testing.
The Agro sisters are identical twins, so you would expect them to have (almost) identical DNA and, therefore, (almost) identical results. But, as you've probably guessed by now, that's not what happened.
The Living DNA results found that Charlsie has DNA from Scotland and Ireland while Carly has a small percentage of ancestry from England.
The 23andMe results suggested Carly is almost 10 percent more "broadly European" than Charlsie. Meanwhile, Charlsie has 2.6 percent French and German ancestry, which Carly, apparently, does not. Charlsie also came back as 3.3 percent more Eastern European, 1 percent more Italian, and 1 percent more Balkan than her sister. Hey, at least, they can agree that the twins are largely of European descent.
Perhaps more puzzling is the differences between brands. AncestryDNA gave the twins mostly Eastern European heritage (38 to 39 percent), MyHeritage mostly Balkan (60.6 percent to 60.7 percent), while FamilyTreeDNA found that between 13 to 14 percent of their DNA came from the Middle East.
Of course, if these tests were 100 percent accurate, they would come back with identical results. So to find out what is going on here, Marketplace had the results from the five companies sent to Yale University for analysis.
According to Mark Gerstein, a computational biologist at Yale University, any results from the same company should give the twins identical results, simply because they would be sending in the same raw data. His research confirmed that this was, indeed, the case – the raw data provided by each twin was statistically identical.
While it is hard to say what accounted for the differences in the Living DNA and 23andMe tests, he suspects the differences between brands stems from the fact that they use different algorithms and reference panels.
When testing a person's DNA, these companies refer to a bank of DNA samples called a reference panel. Each company has their own reference panel comprising of DNA samples from public DNA databases and customers who have already taken the test. (Hence, the more people take the test, the more accurate it should technically be.)
For each person, they take around 700,000 parts of DNA and use an algorithm to compare these DNA parts to those in the reference panel to find their closest matches. This means the algorithm as well as the size, quality, and diversity of the reference panel can affect the accuracy of your results.
That is to say, don't take your results as conclusive, particularly those that claim to be able to trace your ancestry to a particular town or country. Afterall, some home DNA tests can't even tell a human from a dog.