"You share 12 cM with a Bronze Age warrior" is an irresistible sentence, and it is sold under a name — IBD, identity by descent — that promises more than any method can deliver against a genome this old. This article explains what a match with an ancient individual actually is, why we call the product Ancient Matches and describe the result as identity by state, and how to read the numbers without turning a burial into a grandparent.
Two ideas that sound alike#
Identity by state (IBS) means two genomes read the same across a stretch: the same alleles at the same positions, whatever the reason.
Identity by descent (IBD) means two genomes read the same across a stretch because both copies were inherited from one recent common ancestor — the segment is a single piece of DNA that passed down two lines and met again.
Every IBD segment is IBS; the reverse is not true. Two people can carry identical stretches because both inherited them from a common ancestor, or because the alleles in that stretch are simply common in the population both descend from, or by chance. Telling IBD from IBS is what consumer relative matching does between living people, using long segments, dense genotypes and the statistics of recombination over a few generations.
Why it is IBS against an ancient genome#
Against a person who died four thousand years ago, three things break the IBD inference:
- Time. Recombination breaks inherited segments roughly once per centimorgan per generation. Over 150 generations, a segment shared by literal descent from one Bronze Age individual would be vanishingly short and almost always undetectable — and a stretch that is long enough to detect is far more likely to be shared because it is common in a population than because it descended from that one person.
- The ancient genome is incomplete. Most ancient samples have missing data at many positions and were read at low depth, so "identical across this stretch" is inferred from the markers that happen to be covered, not from a full sequence.
- The panel is a sample of the dead. The individual you match is whoever was excavated, sequenced and published — not a relative located by search. A match says you and that person draw on the same ancestral population; it cannot say more.
Methods specifically built to detect IBD in ancient DNA exist — Ringbauer and colleagues (2024) describe one — and they are designed for pairs of ancient individuals who lived at the same time, where descent from a recent common ancestor is possible. Between a living person and a single prehistoric burial, the honest description of a shared stretch is identity by state. That is why the product is not called "ancient IBD".
What Ancient Matches computes#
Your raw file — a chip export, or a whole-genome VCF via the €10 add-on described in Upload a whole-genome VCF — is scanned against every individual in our ancient reference panel, one person at a time, across the 22 autosomes. For each individual the scan finds the stretches where your genotypes and theirs are consistent, and reports:
- total shared length in centimorgans;
- segment count and the longest segment;
- informative markers inside the segments — the positions that could actually have distinguished the two of you — and their density;
- the chromosome painting: where on your own chromosomes each stretch falls;
- the individual's dates, find location and citation, and a map.
The report ranks every match, rolls them up by population — never showing a population total without the number of individuals behind it — and shows the closest 200 in full. Nothing is metered; there is no tier that shows more matches. The price is €29.99, the same whether bought alone or against an existing qpAdm order.
Reading the numbers#
Centimorgans are a length, not a relationship. In living-relative matching, 12 cM might mean a distant cousin. Against an ancient genome, 12 cM of IBS means 12 cM of consistent reading — informative only in comparison with your other matches and with the population roll-up.
Informative markers decide how much a segment means. A 10 cM stretch with 800 informative markers is a real observation; the same stretch with 40 is barely tested, because the ancient sample was hardly read there. This is why marker density is printed on every row and why a denser file — a whole-genome VCF — changes this product more than any other.
The population roll-up is the finding. One match with one individual is an anecdote. Sharing more, and longer, with the individuals of one ancient population than with those of another is evidence about which ancestral populations your genome draws on — the same question qpAdm and Global25 answer with proportions, approached from the other end.
Rank, do not absolutise. Your top match is your top match in this panel; add a thousand newly published genomes and it may not be.
What a match is not#
- Not an ancestor. A shared stretch is evidence that you and that individual draw on the same ancestral population. It is not evidence that the individual was your forebear, and we never describe a match that way.
- Not a relative. There is no genealogical relationship to name, and no degree to estimate. Estimating degrees between living people is a matching question, and even there the signal blurs past about third degree.
- Not a percentage. Matches do not sum to an ancestry breakdown. For proportions, the instruments are qpAdm and Global25.
A worked reading#
Imagine two rows in a report:
Individual A Bronze Age, Hungary total 31 cM 4 segments longest 11 cM markers 2,140
Individual B Iron Age, Italy total 34 cM 2 segments longest 22 cM markers 310
B has the larger total and the longer segment, and A is the more informative match. B's 22 cM stretch was tested at 310 markers — the ancient sample was barely read there, so "consistent" is a weak statement. A's four segments were tested at seven times as many positions. Neither row says anything about descent from A or B; what they say, taken with every other row from the same populations, is which ancestral populations your genome reads most like — and that is the roll-up to look at before any single name.
How it fits with the other analyses#
Think of three lenses on one genome. qpAdm asks what mixture of ancient populations explains my genome, and is that model even admissible? Global25 asks where does my coordinate sit among ancient and modern rows? Ancient Matches asks which particular published individuals does my genome read the same as, and where? The three should agree in outline — the populations your matches roll up to should be the populations your models draw on — and where they disagree, the disagreement is informative. The individuals themselves are browsable on the free Ancient Sample Atlas, and the panel they come from is explained in AADR explained. The terms are in the glossary.
References#
- Browning, S. R. & Browning, B. L. (2012). Identity by descent between distant relatives: detection and applications. Annual Review of Genetics, 46, 617–633.
- Ringbauer, H. et al. (2024). Accurate detection of identity-by-descent segments in human ancient DNA. Nature Genetics, 56, 143–151.
- Mallick, S. et al. (2024). The Allen Ancient DNA Resource (AADR): a curated compendium of ancient human genomes. Scientific Data, 11, 182.
