Genes, Ancestry, and Populations

TL;DR. Some diseases really are much more common in people with particular ancestry, and the reasons are historical: which pathogens their ancestors faced, which small groups they descend from, and what they ate. Sickle cell trait is common in West Africa because it protects against malaria. Tay-Sachs was common among Ashkenazi Jews because of a population bottleneck. Lactose intolerance is the human default and lactase persistence is the recent mutation. These are real, specific, and clinically useful. What is not useful is race, which is a social category with fuzzy, historically shifting boundaries that does not correspond to the genetic structure of our species. Using ancestry as a clue is good medicine. Using race as a biological variable has produced a long list of documented harms, several of which are only now being corrected.

Key takeaways

• Human genetic variation is continuous and clinal, not clustered into races. Roughly 85 to 90 percent of genetic variation exists within any given population group, not between groups.
• Real ancestry-linked risk variants exist and matter: sickle cell, thalassemia, G6PD deficiency, APOL1, Tay-Sachs, BRCA founder mutations, and several drug-response genes.
• Malaria has shaped the human genome more than any other disease, leaving multiple protective red cell variants that cause disease when inherited in double dose.
• Migrant studies settle many arguments: Japanese populations moving to Hawaii and California acquired the heart disease rates of their new country within a generation, which no genetic explanation can account for.
• Race-based clinical algorithms have caused harm. The race coefficient in kidney function estimation was removed in 2021 after evidence that it delayed transplant referral for Black patients; race corrections in lung function testing are being removed for similar reasons.
• Pulse oximeters overestimate oxygen levels in people with darker skin, a device design problem that caused missed hypoxia during the COVID-19 pandemic.

What human genetic variation actually looks like

In short: Most human genetic variation exists within any population rather than between populations, and it changes gradually with geography rather than in blocks.

Humans are a young, closely related species. Modern humans left Africa relatively recently, and every population outside Africa descends from a subset of African genetic diversity, which is why African populations retain more genetic variation than all other populations combined.

Richard Lewontin's 1972 analysis, repeatedly confirmed with far larger datasets since, found that roughly 85 to 90 percent of human genetic variation is found within any single population, and only about 10 to 15 percent distinguishes populations from one another. Pick two random people from the same continent and two from different continents, and the difference between the pairs is small.

Variation is also clinal: allele frequencies change gradually with geography, because people have always moved and mixed along continuous gradients. There is no line you can draw across a map where the genetics changes abruptly. Skin pigmentation, the trait most used to construct racial categories, is controlled by a modest number of genes under strong selection for ultraviolet exposure and vitamin D synthesis, and it varies almost independently of the rest of the genome. Two people with similar skin colour from different continents can be more genetically distant from each other than either is from someone with very different skin colour.

None of this means population differences are absent. Statistical clustering of genomes does recover geographic ancestry, often with fine resolution, and clinically important variants really do differ in frequency between populations. The point is that ancestry is a continuous, multidimensional, and often mixed thing, and race is a coarse social label that approximates it badly.

Don't be confused: "there are no races biologically" does not mean "ancestry is irrelevant to medicine." It means the categories used on forms are the wrong instrument. A patient's genealogy, geographic origin, and, where it matters, specific genotype are informative. A tick-box that lumps a Somali, a Jamaican, and a fourth-generation African American into one category, and every South Asian and East Asian into another, cannot carry that information reliably.

Why malaria wrote itself into the genome

In short: Five red cell variants persist at high frequency because carrying one copy protects against malaria, at the cost of disease in those who inherit two.

The strongest signal of natural selection in recent human evolution comes from malaria, which has probably killed more humans than any other single infectious disease.

This is balanced polymorphism: a variant that is harmful in double dose is maintained at high frequency because it is protective in single dose, as long as the selective pressure persists. The sickle mutation arose independently at least five times in different regions, which is a strong signature of selection.

Duffy negativity deserves a specific clinical note. Because the Duffy protein also affects neutrophil distribution, people who are Duffy-null have lower measured blood neutrophil counts as a normal variant (sometimes called benign ethnic neutropenia, now more accurately Duffy-null associated neutrophil count). Using a single reference range has led to unnecessary bone marrow biopsies, exclusion from clinical trials, and inappropriate withholding of chemotherapy from patients of African ancestry. Laboratory reference ranges built on one population and applied to everyone are a recurring source of this kind of error.

Founder effects and bottlenecks

In short: When a population descends from few ancestors, whatever those founders carried becomes common, which is why carrier screening is offered by ancestry.

When a population descends from a small number of ancestors, whatever variants those founders happened to carry become common in their descendants, regardless of whether they are beneficial.

Consanguinity (marriage between relatives) raises the chance that both parents carry the same recessive variant, and is customary in parts of the Middle East, South Asia, and North Africa. It raises the risk of recessive disorders in offspring roughly two to three fold above baseline, from around 2 to 3 percent to around 4 to 6 percent for serious congenital conditions, which is a real and manageable increase rather than the catastrophic one sometimes implied. Genetic counselling and carrier screening address it directly and are more effective than exhortation.

Diet, environment, and the genes that adapted to them

In short: Lactose intolerance is the human default, and several ancestry-linked variants only matter if you consume the thing they process.

Lactase persistence. All mammals stop producing lactase after weaning; the human default is lactose intolerance, present in most of the world's adults. Mutations that keep the enzyme switched on arose independently at least four times, in Northern Europe and in East African, Middle Eastern, and Arabian pastoralist populations, and spread rapidly wherever dairy animals were kept. It is one of the strongest signals of recent natural selection in humans. Roughly 65 percent of the world's adults have reduced lactase activity, and the rates range from under 10 percent in Northern Europe to over 90 percent in parts of East Asia. Framing this as a disorder is a mistake: the minority condition is the derived one.

ALDH2 deficiency. A variant carried by roughly a third of people of East Asian descent slows breakdown of acetaldehyde, alcohol's toxic intermediate, producing intense facial flushing, nausea, and palpitations. Carriers who drink despite it have a substantially increased risk of oesophageal cancer, because acetaldehyde is a carcinogen and it accumulates. This is clinically actionable information that maps onto ancestry, and it is entirely inert in someone who does not drink.

APOL1. Two variants common in West African ancestry substantially increase the risk of several kidney diseases (Chapter 23). They persist because they confer resistance to the trypanosome causing African sleeping sickness. This is the same evolutionary bargain as sickle cell, discovered a century later, and it is now a drug target.

PNPLA3 raises the risk of fatty liver disease and is more common in people of Indigenous American and Hispanic ancestry, contributing to higher rates of liver disease independent of alcohol and obesity.

Cystic fibrosis is common in European-descended populations, and the reason its carrier frequency reached roughly 1 in 25 is unresolved. Proposed heterozygote advantages include resistance to cholera or typhoid, and none is established. It may simply be drift.

Pharmacogenomics: where ancestry genuinely changes prescribing

In short: Ancestry decides whom to test, and the test result decides the prescription, which is the correct structure and the one race-based algorithms get wrong.

Several drug-response variants differ enough in frequency between populations that testing is recommended by ancestry, and these are the clearest legitimate uses of the information.

Note what these have in common: the ancestry information is used to decide whom to test, and the decision is then made on the test result, not on the ancestry. That is the correct structure, and it is exactly what race-based algorithms get wrong.

Where race-based medicine went wrong

In short: Four documented cases in which an assumed average difference was converted into a correction applied to individuals, always in the direction of less care.

Kidney function. For two decades, equations estimating glomerular filtration rate from blood creatinine included a "race coefficient" that raised the estimated kidney function of anyone classified as Black by roughly 16 percent. The justification was an observed average difference in creatinine levels, attributed to muscle mass, based on limited data and applied as though race were a biological variable. The consequence was systematic: Black patients' kidneys were reported as healthier than they were, which delayed referral for specialist care, transplant waiting list placement, and drug dose adjustment. After sustained argument, a joint task force recommended a race-free equation in 2021, and it has been widely adopted. Modelling studies estimated that the old equation had delayed transplant eligibility for thousands of patients.

Lung function. Spirometry reference equations have historically applied "race correction," scaling down expected lung volumes for Black and Asian patients. The practice traces back to nineteenth-century measurements by Samuel Cartwright and others, explicitly used to argue for the biological inferiority of enslaved people, and it persisted in equipment defaults into the 2020s. Its effect is to make a given measured lung function look normal in a Black patient and abnormal in a white one, which delays diagnosis and reduces occupational compensation eligibility. Professional societies have now recommended moving to race-neutral equations.

Pulse oximetry. Pulse oximeters estimate blood oxygen by shining light through tissue, and melanin absorbs light. A study published in 2020 found that Black patients had nearly three times the frequency of occult hypoxaemia (dangerously low arterial oxygen despite a reassuring oximeter reading) compared with white patients. During the COVID-19 pandemic this meant patients being sent home or not escalated because a device under-detected their hypoxia. This is not a genetic difference; it is a device calibrated on a narrow population, and regulators are now requiring better testing across skin tones.

Pain. Multiple studies have found that Black patients receive less analgesia than white patients presenting with the same conditions, including in children with appendicitis and in adults with fractures and sickle cell crisis. A widely cited 2016 study found that a substantial proportion of medical students and residents endorsed false beliefs about biological differences, including that Black people have thicker skin or less sensitive nerve endings, and that those endorsing such beliefs rated Black patients' pain lower and recommended less adequate treatment.

The pattern across all four: an average difference (real or assumed) was converted into a correction applied to individuals, in a direction that reduced care.

Migration studies, which settle most arguments

In short: When a population moves, its genes travel and its environment does not, and the disease rates follow the environment.

When a population moves, its genes travel and its environment does not. The results are consistent and decisive.

The Ni-Hon-San study followed men of Japanese ancestry in Japan, Honolulu, and San Francisco. Coronary heart disease rates rose progressively along that gradient, approaching those of the US population, while stroke rates showed the opposite pattern. Same ancestry, different diets and environments, different diseases.

Blood pressure in African-descended populations. Hypertension rates in African Americans are among the highest in the world. Rates in West African populations, from whom many are descended, are substantially lower, with Caribbean populations intermediate, tracking a gradient of salt intake, obesity, socioeconomic conditions, and stress exposure rather than ancestry (Chapter 20).

South Asian migrants develop type 2 diabetes and coronary disease at higher rates than their counterparts in South Asia and at lower body weights than European-descended populations, which implicates an interaction between an inherited body composition tendency and a changed food and activity environment rather than either alone.

Japanese and Korean gastric cancer rates fall in migrants to the United States within a generation or two, tracking H. pylori prevalence and diet.

The diversity gap in genomics

In short: Most genetic studies were done in people of European ancestry, which makes risk scores and variant interpretation less accurate for everyone else.

Roughly 80 to 90 percent of participants in genome-wide association studies have been of European ancestry, far out of proportion to the world's population. The consequences are practical, not merely a fairness complaint:

• Polygenic risk scores transfer poorly across ancestries, typically losing much of their predictive power in non-European populations. Deploying them clinically as they stand would widen health inequalities.
• Variants of uncertain significance are reported more often in patients of non-European ancestry, because there is less reference data to classify them against, which has led to misclassification of benign variants as pathogenic and, in documented cases, unnecessary interventions.
• Drug targets discovered in one population may not generalise. Several important discoveries (PCSK9, for example, first identified through variants studied in a diverse cohort) came from populations that had been under-studied.

Efforts to correct this include H3Africa, All of Us in the United States, and Our Future Health in the UK, and the gap remains large.

How to use ancestry well

In short: Six rules, of which the first is to use ancestry to decide what to test for rather than to assume the answer.

A short practical summary of everything above:

1. Ask about ancestry to decide what to test for, not to assume a result. Sickle cell, thalassemia, G6PD, Tay-Sachs, BRCA founder mutations, and pharmacogenomic variants are all reasonable triggers for testing.
2. Test the individual. A genotype is a fact about a person; a population frequency is not.
3. Do not put race into a clinical algorithm unless there is a specific, validated, mechanistic reason, and be suspicious of the reason.
4. Check whether reference ranges apply. Neutrophil counts, lung function, and several laboratory ranges were derived from narrow populations.
5. Remember that socioeconomic and environmental exposures track ancestry categories closely, so an association with "race" is usually an association with the conditions of life, not with biology.
6. When a group has worse outcomes, look for the cause in access, environment, and treatment before genetics. That is where it usually is.

Sources and notes

Within-population variation: Lewontin, Evolutionary Biology, 1972, and subsequent larger-scale confirmations including Rosenberg et al., Science, 2002, which also shows that clustering recovers geographic ancestry. Malaria-driven selection: Kwiatkowski, American Journal of Human Genetics, 2005. Duffy negativity and P. vivax: Miller et al., NEJM, 1976; Duffy-null neutrophil counts: Merz et al. and subsequent reference-range work. Tay-Sachs screening impact: Kaback et al., JAMA, 1993. Consanguinity risk estimates: Bittles and Black, PNAS, 2010. Lactase persistence convergent evolution: Tishkoff et al., Nature Genetics, 2007. APOL1: Genovese et al., Science, 2010. eGFR race coefficient removal: Delgado et al., NKF-ASN Task Force, JASN and AJKD, 2021; transplant impact modelling: Zelnick et al., JAMA Network Open, 2021. Spirometry race correction history and reversal: Braun, Breathing Race into the Machine, 2014; American Thoracic Society statement, 2023. Pulse oximetry: Sjoding et al., NEJM, 2020. False beliefs and pain treatment: Hoffman et al., PNAS, 2016. Ni-Hon-San study: Kagan et al., from the 1970s. GWAS diversity: Martin et al., Nature Genetics, 2019; Sirugo, Williams, and Tishkoff, Cell, 2019.

Open questions. Why cystic fibrosis carrier frequency is so high in European populations is unresolved. How to incorporate genetic ancestry into clinical prediction without reintroducing race is an active methodological problem. The relative contributions of genetics and environment to most observed group differences in disease remain contested, with the evidence generally favouring environment.

Next: of everything in this book, what actually prevents disease. 👉