Gemini got "mysterious" results by checking 5 ancestral search services for DNA

The chief geneticist of a popular ancestral search company acknowledged that this is "part of science, part of art"

On the left - Charles Agro, on the right - her identical twin sister Carly; verification of some of the DNA testing results

One set of identical twins, and two different profiles for ancestors. At least, this follows from the results of one of the largest companies seeking ancestors through DNA.

Last spring, Charles Agro and her twin sister Carly bought their DNA test kits from AncestryDNA, MyHeritage, 23andMe, FamilyTreeDNA and Living DNA, and separately sent their DNA samples to each of the companies for analysis. And, despite the fact that their DNA is almost identical, they have not received similar results from any of them.

In most cases, the same company tracked the ancestors of the twins to the same parts of the world - with different probability. But the results from 23andMe make it clear that each of the twins in the pedigree had an unexpected turn. If you believe them, Charles’s heredity is 10% less “pan-European” than Carly’s. She also has ancestors from France and Germany (2.6%) who her sister does not have.



Also, the twins had different degrees of belonging to ancestors from Eastern Europe - 28% for Charles and 24.7% for Carly. And if Carly had these ancestors associated with Poland, then Charles had not specified their exact origin.

“The fact that they gave different results for you and for your sister, I find it extremely mysterious,” said Mark Gershtein, a computational biologist at Yale University.

"Shockingly similar" DNA twins

The twins sent the results obtained from all five companies for analysis to the Gerstein team. He says that any results obtained from one company by each twin must be identical. And there is a simple reason: the collected data on the sisters DNA is almost identical. “They are shockingly similar,” he said.

A team from Yale University was able to download and analyze the source data used by each of the companies for the calculations. A complete DNA sample consists of three billion parts, but origin verification companies use about 700,000 of them to work with genetic differences.

According to data from 23andMe, 99.6% of the parts were the same, which is why Gerstein and the team were so surprised by the result. They concluded that the baseline data from four other companies were also statistically identical. But at the same time, none of the five companies gave an identical origin to the twins.

“We think that the numbers should have been exactly the same,” said Gerstein. And although he cannot say for sure why these differences arose, he suspects that it is all about the algorithms used by companies for processing the DNA of users. “The whole thing should be in the calculations. These calculations are carried out by different methods. ”

In response to the question why the twins did not get the same results, despite the similarity of their DNA, a representative of the 23andMe site said that even such small differences could lead to the fact that their algorithm would produce slightly different estimates of the probability of origin. Although the company says they approach the development of their tools and reports with scientific rigor, they recognize that the results are a “statistical approximation”.

Differences in the five companies

In FamilyTreeDNA, the twins reported that their ancestors lived in Sicily, in Poland and in Ukraine. However, the results obtained by each of the sisters contain very surprising — and in some cases contradictory — details of the family history.


Charles Agro with her aunt on holiday in Malta last summer. Before using DNA tests, the Agro sisters believed that most of their ancestors lived in Sicily, Ukraine, and Poland

AncestryDNA found that most of the twin ancestors came from Eastern Europe (38% for Carly and 39% for Charles). The MyHeritage test results place most of their ancestors in the Balkan region (60.6% for Carly and 60.7% for Charles).



More surprising results were presented by Living DNA, according to which a small part of Carly’s ancestors lived in England, and Charles’s ancestors lived in Scotland and Ireland.



Another unexpected turn was found in the results from FamilyTreeDNA, which related 13-14% of twin ancestors to the Middle East - much more than other companies, considering that two of them did not find any Middle Eastern roots at all.

Paul Mayer, the lead geneticist at FamilyTreeDNA, recognizes that determining the genetic differences of people from different places is a difficult task. “By itself, the search for frontiers is the leading edge of science, so this task is at the intersection of science and art,” Meyer said in a telephone interview.

How it works

To find out about human ancestors, companies like 23andMe compare a DNA sample to what is called a reference panel. It consists of a set of DNA samples of users who have previously passed the test, and of samples taken from publicly available databases.

Dr. Simon Gravel, a population genetics specialist at McGill University, participating in the 1000 Genomes Project , said that similar companies take about 700,000 DNA segments and use an algorithm that compares them with segments from the reference panel.

“They are looking for coincidences with different regions of the world,” he said. “As a result, they get some assessment of the contribution of that part of the reference panel that more closely matches your DNA, and from this, an assessment of your origin is derived.” Different companies use different panels, so each of them is likely to give one person different results.

AncestryDNA recognized that the size of the reference panel is a major factor. We were informed that the company "is constantly working to improve the scientific approach," and that "a new, larger panel will give users more accurate results."

Where did the differences come from?

As Gravel says, many factors can affect the accuracy of the results obtained from a company engaged in heredity, but the size and quality of the reference panel is particularly important. The larger it is, the more accurate the results, he says. “If you have fewer people to compare, you have to cut corners more actively,” he said. “You also risk losing diversity, the existence of which in a certain region you did not know.”

Another reason for the discrepancies in the results of different companies is the arbitrary approach to dividing the world into regions, says Gravel. “They have to, roughly speaking, take a pencil in their hands and say:“ This will be the region ”. And different companies draw different borders. ”

He also says that the tests are more accurate for people of European descent, as more people with such heredity have passed.

He warns that you should not consider the results as determined. He says that a company can track a person’s heredity using a DNA sample on a certain continent with statistical error, but some more specific information, such as a country or city of origin, will not be reliable.

Lack of supervision

The largest companies that study heredity in DNA have already tested millions of people. MyHeritage, for example, says that it expects to exceed the $ 100 million bar on sales of services this year. But, despite the popularity of such testing, there is no government or professional oversight of this industry that would guarantee the authenticity of the results. Gravel is worried about this situation.

“In science, there is usually a process of conducting expert assessments and access to data and algorithms is open - this is how we guarantee high quality data,” he said. “And in this case there is no access to them, as companies keep the data secret.”

Therefore, says Gravel, consumers should be skeptical about the results of these tests. People need to understand that these tests do not meet standards such as, for example, medical tests. These tests are more like “scientific entertainment,” he said.

As with 23andMe, MyHeritage says their results are "approximate ethnic estimates." When Rafi Mendelssohn, a representative of MyHeritage, was asked why the company offers its results with such certainty - in the video with the test results that the user receives, it says “you belong to” before the user's pedigree - he said that he considers it obvious that users get the estimated results , and that is especially clear to users from North America.

Results may vary.

Whatever results you get, do not get attached to them strongly. They may change. In September, AncestryDNA informed customers about updating ratings with the following message:

"Your DNA does not change, but we now have 13,000 additional reference samples and a new powerful scientific method that allows you to give improved results by ethnicity." The article uses the results obtained on November 6, 2018, when the company has already updated its data. In the updated results were found kinship ties with Russia, Greece, the Balkans and the Baltic countries, which were not mentioned earlier.