A survey of nearly a million whole genomes describes what a typical person’s DNA actually contains — and finds that most people would be flagged as high risk for about three common conditions.

Genetic testing has a presentation problem. A report that names a variant in a person’s DNA implies that finding one is unusual, and the clinical language around it — pathogenic, variant of uncertain significance — reinforces the impression. What has been missing is the denominator: how much variation an ordinary genome contains, so that any particular finding can be placed against it.

A study in Genome Medicine led by the University of Exeter supplies that figure by surveying nearly a million whole-genome sequences and describing what the average one looks like.

The senior author is Professor Caroline Wright of the University of Exeter.

Five million differences

Every genome contains around 5 million variants — positions where one person’s DNA differs from another’s. About 10,000 of those alter the sequence of a gene, with consequences ranging from lactose intolerance to how a person responds to a given drug, and many with no consequence at all.

“Actually, variation is the norm,” Wright said. “Yet most people are unaware of just how much genetic variation is present in every genome.”

The number that reframes clinical genetics is the next one. Each person carries 1 to 3 rare variants that have previously been shown to cause serious genetic disease, plus about 50 rare variants whose clinical relevance is uncertain.

“Importantly, we found that everyone carries a small number of genetic variants linked to disease,” Wright said, “although many of these will not actually cause disease in that individual.”

That gap between carrying a variant and developing the condition is the crux. A variant earns a disease label by being found in patients who have the disease, and that evidence base is assembled from people who came to clinical attention — which systematically overstates how often the variant causes illness. Penetrance, the proportion of carriers who actually develop the condition, is usually estimated from exactly the biased sample that identified the variant in the first place, and when it has been re-estimated in unselected populations it has repeatedly turned out lower, sometimes dramatically so.

Which means finding a known disease variant in a healthy person is not an anomaly to be explained. It is the expected result, one to three times over, in everybody.

High risk for three things

The team also examined polygenic scores — risk estimates built by adding up the small contributions of many common variants — for 28 common conditions including diabetes, cardiovascular disease and schizophrenia.

Most people would be categorised as high genetic risk for around three of them.

This is arithmetic as much as biology, and it is worth working through. With 28 scores and a threshold that flags, say, the top tenth of the distribution, the average person lands in the top tenth for roughly three conditions simply because the scores are largely independent of one another. Being high-risk for something is therefore the normal condition, not a distinguishing finding — which has a direct bearing on consumer genomic reports that list a person’s elevated risks without saying how many elevated risks a typical report contains.

The team did not assess participants’ health status, so many of those flagged as high risk may have no disease at all.

Sixty-three new mutations, and the father’s age

For individuals recruited alongside their parents, the study could count de novo variants — mutations present in the child and in neither parent. There were about 63 per person, arising mostly in the father’s sperm and some in the mother’s eggs, and the count rose by about 1.2 for each year of paternal age.

The paternal skew and the age effect have the same cause. Sperm are produced continuously from dividing stem cells throughout a man’s life, and each division carries a chance of copying error, while a woman’s eggs are formed before her own birth and undergo far fewer divisions. More years of spermatogenesis means more accumulated mutations, at a rate this study puts at roughly one additional mutation per year.

Whose average

The most important limitation is who was sequenced. The data come from two biobanks — UK Biobank and the American All of Us programme — and participants skew toward people of Western European origin.

That makes this an average of a particular population rather than of humanity, and the skew is not incidental to the findings. Which variants are classified as pathogenic depends on which populations have been studied, so a variant common and harmless in an under-sequenced population can carry a disease label derived from a handful of European patients. Polygenic scores are worse still: they are built from associations in one population and are well documented to lose accuracy, sometimes most of it, when applied to another. A figure like “high risk for three conditions” is not transferable to someone whose ancestry is absent from the training data.

Biobank volunteers are also not representative of the populations they are drawn from. UK Biobank participants are healthier, wealthier and less likely to smoke than the UK average, a selection effect documented in the cohort’s own literature — which bears on any estimate of how many disease variants circulate in healthy people, since the comparison group is healthier than the general public.

Nor does a survey of this kind say what any individual should do. Knowing that everyone carries one to three disease-linked variants is useful for interpreting a result, and it does not tell a particular person whether theirs is the one that matters; that is a question for a clinical genetics service with the person’s family history in front of it.

The contribution is a reference point the field has lacked. Clinical genetics has spent two decades cataloguing what can go wrong in a genome without a well-measured account of what is ordinarily there, and a number for the background rate is what allows a finding to be read as signal or as noise.

The paper was published on 22 October 2026. The work was funded by the Medical Research Council, Wellcome and the NIHR Biomedical Research Centre: Exeter. The release includes no conflicts of interest statement, and names no first or lead author.

---

Notes

1. “The average individual genome: a survey of nearly 1 million whole-genome sequences.” Genome Medicine, 2026. DOI: 10.1186/s13073-026-01792-x.
2. University of Exeter. “Average human genome described for first time, with implications for disease risk.” EurekAlert!, 22 October 2026. https://www.eurekalert.org/news-releases/1148699.

Photo by Edward Jenner on Pexels.com