Monday, February 27, 2017

Inside Edition Investigates Reliability of DNA Ethnicity Tests

Erica McGraw and sisters Nicole and Jaclyn tested with 23andMe.Identical triplets should have identical DNA, right? That is what Dr. David Ku of Universal Genetics told Inside Edition. Inside Edition used three sets of identical triplets and one rare set of identical quadruplets to test DNA ethnicity reports from 23andMe, Family Tree DNA, and AncestryDNA. The results are a rude awakening.

(Select any image in this article to see the Inside Edition video on YouTube.)

One McGraw triplet in the 23andMe test had twice the amount of French and German ancestry as another. According to Family Tree DNA, British Isles ancestry in the Maynard triplets ranged from as low as 59% to as high as 70%. The AncestryDNA test returned a range of British ancestry differing by just 2%. The 23andMe results for the Pyfrom Quads, was the best, giving identical ethnicity.

The Maynard triplets tested with Family Tree DNA.So what gives?

First the idea that identical twins have identical DNA has been called into question in recent years. Last July, a scientist writing for the BBC said that mutations occur in DNA fast enough that “between 10 and 100 new mutations per person…occur early enough in embryonic development to be present in most cells in the body.” Keep in mind, however, that there are 3 billion base pairs. (See “Do Identical Twins have Identical Genes?” in Science Focus, The Online Home of BBC Focus Magazine.) A March 2008 article in The American Journal of Human Genetics reported a scientific study showing that DNA in identical twins is not identical. Something called copy-number variations can cause their DNA to diverge in material ways. (For a layman’s discussion of this study, see “Identical Twins' Genes Are Not Identical” in Scientific American.) Can these differences explain the results found by Inside Edition? I don’t know. It has been known for a while that epigenetics cause variations in identical twins, but epigenetics don’t change the underlying DNA.

New Jersey triplets tested with AncestryDNA.Second, it is possible that fraternal twins can look very similar. Particularly when young, it may require a DNA test to distinguish monozygotic (identical) and dizygotic (fraternal) twins/triplets/quadruplets. One of a couple of the triplets in the Inside Edition broadcast look different enough in the broadcast that to me they could conceivably (unintentional pun) be fraternal. Inside Edition did not disclose the estimated relationships found among them.

Third, the testing companies can’t control how carefully test takers follow instructions and avoid contamination. I imagine there are many common substances in our homes, our food, maybe even our water, that have the ability to contaminate a DNA sample. Can such contamination explain the differences found by Inside Edition?

The Pyfrom Quads took the test from 23andMe.Fourth, I have to believe there are processing errors at the laboratory. I don’t follow the industry, so I don’t know what typical error rates are.

Someone with more molecular biological knowledge than I have will have to weigh-in on what might explain the large differences seen among triplets in the Inside Edition test.

For more information, see “How Reliable Are Home DNA Ancestry Tests? Investigation Uses Triplets to Find Out” on the Inside Edition website.


  1. An interesting comparison, would be to have the same individual test, with the same company, multiple times and see if their ethnicity results are exactly the same each time. I'm guessing they wouldn't be. Why?

    I once had it explained to me that some of the surprising ethnicity variability among close family members, can be partially explained by the ethnicity testing process being a process of sampling and averaging. I'm no scientist, but the way I understood it was not all your genes are checked in the process, and the ethnicity percentage you end up with is an average of them then checking segments within the sample many times (thousands maybe) for the ethnicity markers. This explains the "range" estimates on Ancestry's results. If you have a range of 20-50% British ethnicity, but an average of 35%, that means they found 20% one of the times they checked a sample for markers, 50% another time, but all the times together averaged out to be 35%.

    I am guessing each company has their own ethnicity filters, sampling process, and algorithm for averaging the results. This could explain some of the difference from company to company. But I am also curious if all of the triplets/quadruplets are for certain identical. Even among identical twins there are all sorts of varieties of types based on the manner in which the early cell division happens. See basic explanation here.
    And yes the estimated relationships and total Centimorgans, segments, etc shared in each case would have been useful information.

  2. I believe the reason to take the autosomal DNA test is to discover cousins and/or to help adoptees find their birth family. I would bet that the results of all three testing companies showed that the sisters were sisters and that they shared parents (if the parents had been/were tested); what is unreliable about that? I agree with The AncestryInsider and Sara WS that how the testing is done from start to finish and whether the twin are really identical (one of a triplet can be fraternal and the other two identical, etc.) will probably skew results. DNA testing is one tool in our genealogy toolbox to discover our ancestors, not the ONLY tool. It can give clues on where to research and sometimes pretty definitive answers on kinship.

    My Dad was very disappointed that his atDNA showed no German ancestry but lots of UK and Irish ancestry...well, my understanding is that the migrations of the Angles, Saxons and Vikings, etc. made a mish mash of western European weed out just German is often tough if a bunch of those same "German" genes are extent now in the UK. His Y-DNA testing on the other hand has confirmed the paper trail for me of his German ancestry (immigrants from the Palatinate to PA in 1750).

    I think it is disingenuous of Inside Edition to paint all DNA testing with the same brush. DNA testing is a complex science and to reduce or dismiss it based on extremely small sample size is not scientific. And not mention Y or mitochondrial testing at all is poor reporting. Dr Stork hit part of the nail on the head when he says: "I think the answer here is that we've come so far in terms of genetic testing, but you can't just spit in a cup and have every single answer that you are looking for." Folks want all the answers tied up with a bow.

  3. It wouldnt matter if triplets or single each dna holds a combination of both lines of dna that made them my brothers and my sisters and my cousins all share some exact dna but my sisters might have a different dna cms in one line than I do or my brother bottom line if you share same parents you will have a given number from both mother and father but not the same from each that is why it is important to have all siblings test even half siblings can help when working out the geno tree.

  4. I bristle when I see the commercials, which tout the test as a way to establish one's ethnic identity. Balderdash! The testing has little to do with ethnicity. This came to light for me after comparing my results with my sister: she was 56% Great Britain and I was 5% After much research and listening to experts, I discovered that there is very, very little difference between some of the categories, such as Great Britain and Western Europe. In fact, Great Britain does not even mean Great Britain; it means Anglo Saxon, which of course can show up in multiple countries and geographies. Why label the category as Great Britain? Well, the "marketing" answer is obvious: who wants to be labeled Anglo Saxon? And, then, there is the "Irish" category. It does not really mean Irish; instead, it means "Celtic," which includes, of course, numerous geographies and countries, including Scotland, Wales, Great Britain, France, etc. So, why label it Irish? I don't know for a fact, but there is something really cool about being Irish these days, so the marketing makes sense. It is not that the categories are arbitrary; they are not, as our genetic makeup can differ, even radically, from sibling to sibling. The point is that the categories, especially the percentages in the categories, are relatively meaningless at least with respect to ethnicity. My sister and I may have completely different DNA, but we have the exact same ethnicity. So, what does the difference in DNA mean? Nothing. The DNA tests are helpful in establishing cousins for further research, but I have found that unless you really work the system like my family history superstar son, these connections can be a bit cumbersome to work out. The bottom line is that DNA testing is useful for medical research and understanding migration patters, but not much else.