Radio Review: Crowdscience - Can I trust DNA ancestry tests?

'I just took a DNA test', sings Lizzo on her 2017 hit single 'Truth Hurts'. Her results seem fairly conclusive – but exactly how accurate are DNA heritage tests? That is the subject of this episode of CrowdScience from the BBC's World Service.

The episode was not however inspired by Lizzo, but by a question emailed in by listener Karen. There are currently a number of companies offering DNA heritage tests, usually promising to break down your ethnicity into percentages attributed to different geographical regions.

Karen's test initially told her she was 39 percent Scandinavian, but by the third update she was down to only two percent Norwegian. Presenter Dr Alex Lathbridge sets out to investigate. He first checks his own 23andMe results from nine years ago (a Christmas present), and finds that his percentages have not changed, but the regions have become more specific.

He calls Dr Joanna Mountain, senior director of research at 23andMe. She explains that they try to ensure accuracy by using big databases of reference samples from around the world. At 23andMe, over 14,000 people (mostly customers) were chosen, based on their survey answers and genes, to reflect populations that existed before transcontinental travel (at least 500 years ago). Their DNA makes up the reference database. The database only consists of modern DNA and is therefore built on an underlying assumption that people have not moved around, so that 'that the people we are examining today are representative of their ancestors'. This is why, Dr Mountain says, the companies call these component percentages 'estimates'.

So why did Dr Lathbridge's results get more detailed, and Karen's percentages change? Dr Mountain explains that companies are adding more and more reference samples, as well as using new technology. The bigger the database, the more accurate and detailed the results.

Dr Lathbridge points out another problem with the reference databases – they are skewed towards people with European ancestry and have fewer samples from people with African or South Asian ancestry. This is obvious in Dr Lathbridge's own results, where the breakdown of the European ancestry is more detailed than that for African. Dr Mountain claims that her company is trying to address this with initiatives to gather samples from less-represented areas of the world.

The databases are expanding and the algorithms are improving, but will these DNA tests ever be truly accurate, Dr Lathbridge asked Dr Adam Rutherford (broadcaster, author and geneticist). Dr Rutherford explains that DNA tests are good at identifying close relatives, but once you get to the third cousin, their accuracy drops off a cliff. Plus, there is another problem – every time an egg and sperm cell combine, 50 percent of the DNA of each parent is lost. Those losses accumulate throughout the generations, so by the time you've gone back 11 generations, you have absolutely no DNA in common with half of your actual ancestors. Your actual family tree and genetic family tree diverge.

So, to answer Karen's question, these tests can never be completely accurate at describing your ancestry – no matter how good the technology, and how extensive the reference database.

Debbie Kennet, a genetic genealogist and family historian, stresses that documents (eg birth certificates) are more useful in figuring out genealogy than DNA tests. She says that DNA tests are a last resort for breaking through 'brick walls', where nothing is known about an ancestral line.

The programme goes on to explore a specific instance of this – where a paper trail does not exist or has been erased. This was what happened to the descendants of people taken from Africa during the transatlantic slave trade. Another listener email inspired this topic. Anna asked if someone with her West African heritage would be genetically similar to someone from Haiti. The modern Haitian population is Afro-Haitian, as many people's ancestors were taken from West Africa hundreds of years ago. She asks how it is possible that so much genetic time has passed, while she still sees so many similarities between the two groups.

Dr Lathbridge calls Dr Rutherford back in, who has recently authored a book on race and genetics. Surprisingly, he claims that the chances of the two groups retaining genetic similarity would be low. This is due to the 'forcible mix up' which occurred as 12 million people from multiple countries were imprisoned on the same boats. Dr Lathbridge wonders if DNA tests could be used to cast some light, especially as the written record is so sparse in most cases.

One DNA company is targeting this market of African consumers. It is 'the missing component of our identities' says Dr Gina Paige, co-founder of the company African Ancestry. They use different techniques to 23andMe, which looks at DNA from all over the genome. Instead, their unusual method of DNA analysis focuses on building family trees, going back 500 to 2000 years through the maternal line (using mitochondrial DNA). They can even do the same with a paternal line (using Y chromosomes) – this technique is less common, and only works for men. African Ancestry compares these conserved sequences to people living all over the African continent. 'When you know nothing about any of your ancestry, any piece you can find is empowering', says Dr Paige.

Others are more sceptical of this approach. Genetic genealogist Shannon Christmas (who used to be an unpaid 23andMe ambassador) says that a third of African American men do not have Y chromosomes that trace back to Africa (instead, being European or Native American). He says a better approach is to use classic DNA tests to find living relatives on the African continent. Christmas also argues that identity is shaped by lived experience, not genetics (an unusual perspective for a genetic genealogist). He cautions that we should be careful of appropriating cultures that are 'not our own'. This reminds me of UK listener Karen, who is mourning the loss of her Scandinavian identity. The question of what constitutes an identity is an interesting one, and I would have liked the episode to explore it further.

Returning to the question of how accurate these tests are, it seems this depends on what you are trying to find out. In any case, it is clear that they will never be able to tell you everything about your ancestry.

The programme also ends with a warning – that in taking a test you are sharing 'valuable private data about yourself', and you should weigh up whether or not it is worth it. In my opinion this privacy concern warrants more than a sentence.

I was also disappointed that this programme did not explore any criticisms of population science, for example the work of Angela Saini (author of Superior: The Return of Race Science). She calls population categories 'imaginary, arbitrary', and explores the unethical history of the population genetics field. It also struck me that three of the five experts interviewed for the episode are affiliated with DNA testing companies, which may have resulted in this slightly favourable bias. However, a good amount of science filled the 33-minute episode, and the human stories made this a nicely-balanced account. I would recommend this to anyone looking to learn more about DNA ancestry tests.