What an ancestry DNA test can and cannot tell
An ancestry DNA test compares genetic markers with reference samples from living populations. Its estimates depend on the samples and methods used, and sharing the data can reveal information about relatives as well as the person tested.
What the comparison means
Ancestry DNA testing uses small variations in the human genome, including single nucleotide polymorphisms (SNPs), which are variations at individual nucleotide positions.
Ancestry-informative markers are SNPs that occur at markedly different frequencies in different geographic populations. Testing companies compare a consumer's markers with proprietary reference databases drawn from people living in different geographic regions and estimate the proportions matching variations frequent in each region. For example, if 60% of a consumer’s DNA has variations most frequent in Germany, there is a good chance that those regions of DNA came from ancestors who lived there. A variation common in one region can also occur elsewhere: one frequent in German reference samples may also occur in samples from Japan or Italy. Such matches support probabilistic estimates of ancestry rather than a definitive account of where every ancestor lived.
Genetic genealogy can show where people sharing some of a consumer's DNA live today, but it cannot determine where all that consumer's ancestors lived, which race or social groups those ancestors identified with, or how the consumer should identify today.
Why estimates differ
The quality and size of the reference database strongly influence ancestry estimates. Companies try to select reference samples from people whose families have lived in a region for several generations. In 2018, little was publicly known about sample inclusion and exclusion requirements beyond family documentation and the requirement that reference individuals be unrelated.
Geographic representation can be uneven. An AncestryDNA reference panel described in 2018 contained more than 16,000 samples, including 2,072 from Germanic Europe, 1,959 from Eastern Europe and Russia, and 1,519 from England, Wales and Northwestern Europe, compared with 41 from Northern Africa, 31 from Senegal and 30 from Sardinia. Nonrandom sampling could introduce bias.
Ancestry predictions require complex algorithms and statistical analyses; different analytical methods applied to the same reference database would likely produce different predictions. A company's results may change as its database and algorithms are updated.
Information beyond ancestry
In 2018, ancestry DNA tests had little oversight by scientific associations and were not regulated by the US Food and Drug Administration, whereas the agency regulated the marketing of direct-to-consumer tests for genetic predisposition to certain diseases or conditions.
Although ancestry tests make no health-related claims, the SNPs they use may lie in coding or noncoding regions of genes. Studies have linked polymorphisms with disease susceptibility, illness severity and response to drug treatment, and research may discover disease associations for SNPs with no currently known link.
Relative-finding services may reunite estranged relatives who want contact, but they may also reveal infidelity or previously unknown children and cause lasting family conflict. In 2018, few companies warned customers that relatives might not want these discoveries or that making them public could harm relationships. An untested person can therefore be affected by a relative's decision to take an ancestry test.
Data use and privacy
Ancestry testing companies hold distinctive genetic data, and precautions do not eliminate the risk of hacking or database breaches. Anonymising genetic data is difficult: a highly skilled researcher with some knowledge of a family tree might theoretically identify the person behind a record. Withdrawing from research or data sharing after information has gone to third parties may make it difficult to ensure that every shared copy is destroyed.
In 2018, many ancestry testing companies sold consumer data to pharmaceutical companies for research and product development. At that time, 23andMe offered an opt-in choice for research by academic, nonprofit and industry organisations; examples included Genentech's Parkinson disease research and Pfizer's lupus and inflammatory bowel disease research. Contracts often relinquished consumers' rights to profits from products developed using their genetic information and reserved companies' rights to change terms without notification or consent. In 2018, most ancestry testing companies emphasised anonymisation and opt-in sharing with third parties, and commonly allowed withdrawal from sharing or research and requests for permanent deletion. Consumers often had to assess companies’ data-security and breach-disclosure policies themselves, while contracts commonly waived liability for information theft.
In 2018, GEDmatch offered a free online genealogy database in which users could upload DNA sequences and family trees for comparison with other users’ records. Genealogy databases can also enable criminal investigations. In 2018, companies could decide whether to cooperate with law enforcement requests in the absence of a court order; several leading companies, including 23andMe, stated that they resisted requests for personal information as far as possible. At that time, police could access the publicly available GEDmatch database without a court order. In the 2018 Golden State Killer investigation, police used GEDmatch data to narrow possible suspects to one family, then tested DNA from discarded rubbish to identify a retired law enforcement officer as the suspect.
Consumer understanding of ancestry testing needs to include scientific accuracy and privacy implications as well as its recreational appeal.