Ancestry: how the estimate is built, and why it moves
The reason most people took a DNA test in the first place — and the category where the gap between how results are presented and what they actually are is widest.
How is ancestry worked out from a DNA file?
By comparison, not measurement. Your markers are compared against a reference panel — people whose origins are known — and the result is an estimate of how much of your genome resembles each group. We use the open 1000 Genomes and HGDP panels with open-source methods, so the comparison can be inspected rather than taken on trust.
That word "estimate" is not hedging. There is no place in your genome that records a country. What exists is a pattern of variation that is more common in some populations than others, and a statistical judgement about how well your pattern matches each reference group.
What does an ancestry breakdown include?
Three separate things: a continental and subcontinental admixture breakdown, your mitochondrial and — where applicable — Y-chromosome haplogroups, and an estimate of how much of your genome is Neanderthal. They answer quite different questions, rest on different methods, and are not equally reliable.
- Admixture breakdown
- Continental and subcontinental, computed against the open 1000 Genomes and HGDP reference panels.
- Haplogroups
- Your mitochondrial haplogroup traces the maternal line — mother, her mother, and so on — and the Y haplogroup the paternal line, where a Y chromosome is present in the file. Each is precise about one line and silent about the rest of the family tree.
- Neanderthal fraction
- How much of your genome traces to Neanderthal ancestry, as a proportion of the modern human genome.
Haplogroups trace one line each — mtDNA follows your mother's mother's mother, the Y chromosome your father's father's father. They are precise about those single lines and silent about everyone else in your family tree, which is most of it.
Why do two DNA services give me different ancestry results?
Because each builds its own reference panel, draws its own region boundaries and runs its own algorithm. Change any of those and the percentages change. Two services can read the identical file, report meaningfully different results, and both be defensible — and that is also why your results shift when a company updates its panel.
Continental-level estimates and haplogroups tend to agree across services. Fine-grained regional percentages tend not to — and those are the ones marketing puts on the box.
More on why services disagree →
The other categories
- Medicines How inherited differences change the way people process common medicines, which genes a consumer DNA file covers, and the two that it cannot read reliably at all.
- Family planning What being a carrier means, which recessive conditions a consumer DNA file covers, and why a clear result is not an all-clear.
- Nutrition Which nutrition and wellness markers a consumer DNA file covers, how strong the evidence behind them is, and why a blood test still beats a genotype.
- Fitness What a consumer DNA file can suggest about power versus endurance, response to training, injury predisposition and recovery — and the limits of all four.
- Traits Which everyday traits can be read from a consumer DNA file, which of them rest on strong single-gene evidence, and which are much weaker than they look.