Like many Minnesotans, Jeff Weiss had a vague understanding that his ancestors came from Scandinavia and Germany before coming to the U.S.
“Everything we knew was mainly Germanic and Swedish-type people,” Weiss said, having come to the U.S. from Sweden, Germany, Poland and thereabouts. Other than that, the details of his family’s provenance, as far as he knew, were murky.
Weiss, a Shoreview resident, decided to take one of the many DNA tests now on the market that claim to be able to estimate family ancestry by analyzing sequences of customers’ DNA, comparing it to the DNA in their databases and apportioning origin based on similarity.
So did his identical twin brother, Jay, who lives in Portland, Maine. Both being engineers with an interest in science, the brothers turned their quest to learn more about their family history into an experiment.
As identical twins, their DNA is nearly the same. So shouldn’t two companies’ analysis of their heritage match?
There are genetic testing companies that advertise a specialty in uncovering all sorts of different things hidden in people’s genetic material: from estimating their propensity to get heritable diseases to finding out ancestors’ origins and lost relatives.
The brothers did some research and selected two separate companies that seemed to have good reputations for the ancestry part.
“I have enough relatives I don’t get enough chance to see anyway, so I really wasn’t interested in finding more,” Jeff said. He took a test from FamilyTreeDNA. Jay took a test from MyHeritage.
It turned out to be an interesting experiment, one that raised more questions than it answered.
The companies estimated each of the brothers’ genes came from 100 percent from European sources, but the top line is pretty much where the similarities in their results ended.
Jeff’s estimates came back more than three quarters British Isles origin. Jay’s came back with less from that region, but nearly two-thirds Scandinavian. Jay’s estimate included Iberian genes; Jeff’s did not. And the twins got fairly different results for Southern Europe.
How it works
Companies that analyze ancestry based on heritage generally work like this: The company processes a sample of the customer’s DNA, looking for similarities between it and a baseline sample of people with ties to a specific geographic area, say, Europe, South America or Africa. Using a computer algorithm that has assigned certain traits to specific populations, it spits out an estimated breakdown.
Jeff and Jay’s ancestors were the same people, so in theory, their results should be the same. But there are several possible explanations for why they weren’t, said Heather Zierhut, an assistant professor of genetics, cell biology and development at the University of Minnesota. There’s also the possibility of human error, of course. these companies process lots of samples and it’s always possible something got mixed up.
But assuming nothing went wrong in that department, it’s likely that the two companies look at different parts of the human genome. DNA analysis companies typically look at a small subset of the human genome. If they’re looking at different small parts, they could get different results.
Another factor could be differences in the populations the two companies use as a baseline to define different geographic origins.
“How they identify whether someone is coming from the British Isles or England or Greece or Southern Italy from the ones that you had can be different because who they’re comparing to is different,” Zierhut said.
Rafi Mendelsohn, the director of public relations and social media at MyHeritage, said his company DNA tested thousands of people who have a connection to an ethnicity or geographic area to create a genetic baseline.
FamilyTreeDNA uses its own customer database and data from several genetic databases, including the Human Genome Diversity Project, the International HapMap Project, the Estonian Biocentre and the 1000 Genomes Project, wrote company population geneticist Paul Maier, in an email.
How the companies decide what percent of each sample comes from disparate geographic areas can differ too.
“Over time, the lines of where those countries are have changed, so when you think about where our relatives came from, they kind of have moved and migrated as well,” Zierhut said. For example: how each company defines British Isles ancestry, given that the Romans were in the area (and a lot of other places) more than 2,000 years ago.
MyHeritage says it can provide estimates for 42 different ethnicities, while FamilyTreeDNA says it estimates ancestry from 24 populations.
If differences in methodology are likely explanations, what would have happened if Jeff and Jay had taken tests from the same company?
They likely would have gotten more similar results, but discrepancies wouldn’t be unheard of.
Identical twins aren’t exactly identical, despite starting as the same egg before splitting. Mutations at certain parts of the genetic sequence, or ones that happen during gene copying processes can occur early in development, after eggs have split, Maier wrote. It’s also possible the analysis came through more clearly for one twin on a certain gene than another, which caused the algorithm to assign ancestry to one group versus another.
And what about siblings who aren’t twins?
If they take tests from the same company, they should expect estimates to be similar, but not exactly so.
Mendelsohn uses a chicken soup analogy to explain why. If you make a chicken soup with chicken, vegetables and spices and ladle out two bowls, each bowl will have slightly different amounts of chicken and vegetables. So, too, with siblings. One might have their grandmother’s nose, and another a grandfather’s on the other side of the family.
Because of all these factors, results should be taken as estimates of ancestry — not precise breakdowns, Mendelsohn said.
Jeff was disappointed after getting the results back. He knew that migration patterns make it tough to get precise results that pinpoint, say, what town a test-taker’s family came from. But he feels frustrated with the way the kits are marketed, when he felt he got a scattershot estimate at best.
Being twins, the two were pretty sure it wasn’t one of those cases where two siblings take tests and find out they’re not related, an uncommon, but not unheard-of result as DNA kits become increasingly popular.
But the experience left Jeff frustrated and not entirely sure his kit hadn’t been mixed up with somebody else’s.
“(Jay’s) came back a lot closer to what we would have expected it to look like,” he said.