Archaic ancestry
Neanderthal DNA in the file you already have
Aimosti reads 9 named segments of DNA that modern humans inherited from Neanderthals, each tied to the study that found it. A whole-genome gVCF can be read for all 9. A chip export from 23andMe, AncestryDNA or MyHeritage can be read for at most 4, and the real exports we measured read between 1 and 4. We report which ones you carry as a count, never as a percentage of your genome, because a list this short cannot support a genome-wide figure and no one can check such a figure for a single person.
What each kind of file can read
Most of these segments are marked by a single letter at a single position. A genotyping chip reads only the positions it was built to read, so it sees a few of them. A sequenced genome sees them all, as long as the file records where it looked.
| Your file | Segments read, of 9 |
|---|---|
| A chip export23andMe, AncestryDNA, MyHeritage, FamilyTreeDNA, Living DNA | Up to 4, depending on the array |
| A whole-genome gVCF | All 9 |
| A whole-genome plain VCF | Only the ones you carry |
| Aligned reads, BAM or CRAMthrough Deep Read | Whatever the genome file beside them reads |
The other 5 segments are marked only at positions that no consumer array we measured carries, or that cannot be read reliably from one. When your chip does not carry a marker, the report names it and leaves it out of the count, since your array never looked at that position.
A plain VCF lists only the positions where you differ from the reference genome, and every Neanderthal variant on this panel is the non-reference letter. So a plain VCF shows the segments you carry and says nothing about the rest, and the report does not guess whether that silence means “read, and not carried” or “never read”. A gVCF records the positions it examined, which settles it. Deep Read adds a pharmacogene panel from your aligned reads on top of a genome report; this section comes from the genome file.
Measured on real chip exports
One real export per chip version, read by the same code that reads yours:
| Export | Segments read, of 9 |
|---|---|
| 23andMe v3, 2014 export | 3 |
| 23andMe v4, 2020 export | 3 |
| 23andMe v5, 2019 export | 1 |
| AncestryDNA, 2018 export | 3 |
| AncestryDNA, 2024 export | 4 |
| MyHeritage, 2018 export | 3 |
| FamilyTreeDNA, 2020 export | 1 |
| FamilyTreeDNA, 2023 export | 3 |
| Living DNA, 2019 export | 3 |
Two exports of the same brand can differ, because vendors revise their arrays. Your own report counts from your own file’s list of positions, not from this table. Not sure which file you have? Supported files explains each one, and what we analyse lists every module by file type.
The 9 segments, and what each is linked to
Each segment below was traced to Neanderthals in a published study, and each marker was checked against sequenced Neanderthal genomes before it went on the panel. The module can also hold Denisovan segments; none is on the panel yet. The strength label is about the link to a trait, which is a separate question from whether the segment is archaic. These are associations measured across large groups. None of them predicts anything about one person, and none is a diagnosis.
-
An evening-preference segment near ASB1
Several Neanderthal segments cluster in sleep and body-clock traits. This one shows up when hundreds of thousands of people are asked whether they are a morning or an evening person.
What the study found. The archaic allele is associated with answering the morning-or-evening question as an evening person, at p = 3.6 × 10⁻¹⁰. The genome-wide analysis puts the size of it at roughly a 9% shift in the odds of that answer, which is a small nudge inside a question people answer differently in different weeks of their lives.
What it isn’t. Not a sleep finding, not a measurement of your body clock, and not an explanation of when you fall asleep. It is one variant among many that nudge the same self-report.
- Dannemann and Kelso 2017, American Journal of Human Genetics: the contribution of Neanderthals to phenotypic variation in modern humans (PMID 28985494)
- Jones et al. 2016, PLoS Genetics: genome-wide association analyses in 128,266 individuals identify new morningness and sleep duration loci (PMID 27494321)
- dbSNP, rs75804782
-
A pigmentation segment near BNC2
Neanderthal DNA turns up repeatedly around the genes that set skin and hair colour. This is one of those segments, sitting just past the end of BNC2.
What the study found. They found this allele associated with self-reported skin colour, in the direction of slightly darker skin, at p = 1.6 × 10⁻¹⁴. The measurement is a seven-category questionnaire answer, and the shift is a fraction of one category.
What it isn’t. Not a prediction of your skin colour and not a skin-health finding. Pigmentation is set by many genes together and then by how much sun a person has met.
-
A hair colour segment near MC1R
Of every Neanderthal allele tested against 112,000 people, this one produced the strongest association with any trait, and the trait was hair colour.
What the study found. The association with natural hair colour before greying reached p = 3.7 × 10⁻²⁰², and carriers were under-represented among people reporting red hair. A p-value that size reflects how cleanly hair colour can be recorded across a hundred thousand people, not how much of any one person's hair colour this explains.
What it isn’t. Not a hair colour prediction and not a test for red hair. What it reports is that one segment beside a pigmentation gene came from a Neanderthal.
-
The chromosome 3 COVID-19 segment
In 2020 the largest genetic study of severe COVID-19 found its strongest common signal on chromosome 3. The segment carrying it turned out to have been inherited from Neanderthals.
What the study found. In that meta-analysis this allele carried odds of about 1.9 for critical illness, 1.6 for hospitalisation and 1.2 for testing positive at all. Those are ratios rather than risks, and they were measured against a backdrop where age moved the absolute outcome by a factor of hundreds.
What it isn’t. Not a susceptibility test, not a measure of how you would fare against an infection, and not a statement about your immune system. It reports where one segment of your chromosome 3 came from.
-
A drug-metabolising enzyme pair
One of the better-characterised reduced-function haplotypes in clinical pharmacogenomics sits on a stretch of chromosome 10 that modern humans inherited from Neanderthals. The two genes lie about 300 kb apart and the archaic segment spans both, so their variants travel together.
What the study found. This stretch carries the variants known as CYP2C9*2 and CYP2C8*3, which reduce how quickly those two enzymes clear some of the substances they act on.
What it isn’t. Not a medication finding and not a dosing statement. What these variants mean for prescribing is reported separately, in the pharmacogenomics part of your report; this card only says where the segment came from.
-
An EXOC2 segment
One of the weaker signals in the UK Biobank survey of Neanderthal DNA, kept in this panel because its provenance is solid even where its consequence is thin.
What the study found. The archaic allele was associated with natural hair colour before greying at p = 2.9 × 10⁻⁹. That clears the genome-wide threshold and is the weakest association of the nine loci in this panel; the paper does not single out a shade for it.
What it isn’t. Not a prediction of your hair colour. Hair colour is mostly set by variants at MC1R and a handful of other genes, and this is not one of them.
-
Growth hormone receptor
Neanderthals carried a growth hormone receptor with two amino-acid changes and a deleted exon. The deletion is shared with Denisovans, and the two coding changes are found in all three high-coverage Neanderthal genomes sequenced so far. Some living people inherited this receptor through interbreeding.
What the study found. Biobank analyses across more than a million people put the effect at roughly 3 mm of height and 285 g of body mass per copy. Statistically clear, replicated, and far too small for anyone to notice in themselves.
What it isn’t. This is a fact about your ancestry, not a fact about your body. It is not a measurement of your height or build, not a health finding, and not a prediction of anything.
-
The SLC16A11 transporter segment
A Neanderthal-derived version of a cellular transporter is one of the more consequential archaic segments known, and it is common in the Americas and rare in Europe.
What the study found. The haplotype was associated with type 2 diabetes at an odds ratio of 1.29 in the discovery cohort and 1.20 on replication. The association was stronger in people who were younger and leaner at diagnosis. What made it matter in Mexico is how common it is there rather than how strong it is.
What it isn’t. Not a diabetes test and not a risk estimate for you. It reports where one segment of chromosome 17 came from, and a single variant is not how diabetes risk is assessed.
-
The STAT2 segment
A quarter of a megabase around STAT2, an interferon-signalling gene, matches Neanderthal sequence. It is uncommon across Eurasia and common in Melanesia, and it was the first case where introgression looked as though it had been favoured rather than merely tolerated.
What the study found. The haplotype sits at roughly 5% across Eurasia and at about 54% in Melanesian populations, a tenfold difference. A neutrality test that accounts for population history rejected drift alone as the explanation, which makes this a candidate for an archaic variant that was actively favoured somewhere.
What it isn’t. Not an immune finding and not a disease result. The selection signal was found in Melanesia and says nothing about any individual anywhere.
Why we don’t give you a Neanderthal percentage
A genome-wide Neanderthal figure is a sum. To get one, the genome is phased into its two copies, a statistical model compares each stretch against archaic genomes and against African reference genomes, and the length of everything the model calls archaic is added up. We don’t run that model. It needs phasing we do not do, it degrades on the roughly 600,000 positions a chip reads, and the number it produces cannot be checked for any one person, because there is no second measurement to hold it against.
We started from the other end: a short list of segments, each named in its own paper and read at a verified position. You can check every line of that. It cannot be summed into a share of your genome, though. With 9 segments, a percentage would claim a precision nobody has, and a percentile would rank people who mostly tie.
What the report gives you instead is the count with its denominator (how many you carry, out of how many your file could be read for), which archaic population each came from, and how your count compares with what someone of your ancestry would typically carry. That last figure is worked out from population frequencies, and it comes with its main caveat printed beside it: it treats the segments as independent, and they aren’t quite.
What it looks like in a report
This is the top of the section in the sample report, rendered by the same template your report uses, with one of its cards. The sample is a made-up whole genome read from a gVCF, so all 9 segments could be read, and it carries 2 of them.
Archaic ancestry
Exploratory · the call is solid, the meaning is notYou carry 2 of the 9 archaic haplotypes that could be read from your file.
These are named, individually published segments that modern humans inherited from Neanderthals or Denisovans. This is a count over that named set, not a percentage of your genome. We don't report one, because with a set this size a percentage would imply a precision nobody has.
By source: 2 Neanderthal.
Someone with your ancestry would average 1.1 of these. A count of 2 or more happens about 29 times in 100. This treats the markers as independent, and they aren't quite. Someone with more total archaic ancestry is likelier to carry each one, so the real spread is wider than this figure suggests.
chr3p21.31 (LZTFL1 region) The chromosome 3 COVID-19 segment carried Replicated · small effect
You carry one copy of this archaic variant.
- Your genotype
- C/T (read at rs10490770)
- Inherited from
- Neanderthal
In 2020 the largest genetic study of severe COVID-19 found its strongest common signal on chromosome 3. The segment carrying it turned out to have been inherited from Neanderthals.
How seriously to take this: the evidence
- Strength
- Replicated · small effect
- Source of the claim
- Zeberg and Pääbo, Nature, published online 30 September 2020 (PMID 32998156), traced the segment to Neanderthals. The risk estimates come from the COVID-19 Host Genetics Initiative's meta-analysis of up to 49,562 patients across 46 studies in 19 countries (Nature, published online 8 July 2021, PMID 34237774).
- Effect
- In that meta-analysis this allele carried odds of about 1.9 for critical illness, 1.6 for hospitalisation and 1.2 for testing positive at all. Those are ratios rather than risks, and they were measured against a backdrop where age moved the absolute outcome by a factor of hundreds.
The segment is about 50 kb long and sits beside genes involved in immune signalling, including LZTFL1 and a cluster of chemokine receptors. Which variant inside it does the work has never been settled. Its distribution traces migration rather than biology: common in south Asia, present in Europe, near absent in east Asia and in Africa.
This is one of the most heavily replicated associations in human genetics, and almost all of it was measured on unvaccinated people meeting the virus as it was in 2020 and 2021. What the same segment means for someone now, after vaccination and after the virus changed, is a different question and this data cannot answer it.
What this isn’t: Not a susceptibility test, not a measure of how you would fare against an infection, and not a statement about your immune system. It reports where one segment of your chromosome 3 came from.
Zeberg and Pääbo 2020, Nature — the major genetic risk factor for severe COVID-19 is inherited from Neanderthals (PMID 32998156) · COVID-19 Host Genetics Initiative 2021, Nature — mapping the human genetic architecture of COVID-19 (PMID 34237774) · dbSNP, rs10490770
Common questions
How much Neanderthal DNA can you read from my 23andMe raw data?
Some of it. A 23andMe export can be read for at most 4 of the 9 segments on our panel, and how many depends on the chip version. On real exports we measured, a v5 file read 1, a v4 file 3 and a v3 file 3. The report tells you which of those you carry. It does not turn them into a percentage of your genome.
Can a DNA test tell me what percentage Neanderthal I am?
A genome-wide percentage is a model's estimate: the genome is phased into its two copies, compared stretch by stretch against archaic and African reference genomes, and the stretches called archaic are added up. We do not run that model, and nothing in your file lets anyone check the result for you personally. What we read is a short list of named segments, each at a verified position, reported as a count.
Which file reads the most Neanderthal segments?
A whole-genome gVCF reads all 9, because it records the positions it examined as well as the ones where you differ from the reference. A plain whole-genome VCF shows only the segments you carry. A chip export reads between 1 and 4 on the real exports we measured, and never more than 4.
Is carrying a Neanderthal segment bad for my health?
None of this is a diagnosis. Some segments have published links to disease risk, measured as modest changes in odds across large groups of people. Each one is reported at the strength of its evidence, with a line on what it does not mean for you, and the clinical meaning of the drug-metabolism segment lives in the pharmacogenomics section under a pinned guideline.
Do you report Denisovan DNA?
Not yet. The module has room for Denisovan segments, and the 9 on the panel today all trace to Neanderthals.
For the other half of the ancestry section, the maternal and paternal lines, see what a haplogroup is.