Researchers analyzed the DNA of 72,635 living people and found 168 genetic variants actively being eliminated from the human population. Several of them cause cancer.
The idea that human evolution has essentially stopped is one of those assumptions that feels intuitive without being rigorously tested. Modern medicine keeps people alive who might not have survived in earlier eras. Reproductive technology allows people with genetic conditions to have children. The environmental pressures that shaped ancient human genetics, famine, predation, infectious disease, have been dramatically reduced in many parts of the world. It is reasonable to assume that natural selection, at least the version that eliminated disease-causing genes from ancestral populations, no longer operates with meaningful force.
A new study published in the American Journal of Human Genetics has directly tested that assumption using one of the largest genomic datasets ever analyzed for this purpose. The answer is that the assumption is wrong.
Researchers at National Taiwan University, led by first author Jun-Lin Chen and corresponding author Chun-Houh Chen, analyzed 509,817 genome-wide genetic variants from 72,635 Han Taiwanese adults enrolled in the Taiwan Biobank. The participants spanned ages 24 to 70. The core method was straightforward: if natural selection is actively eliminating a harmful genetic variant from the population, that variant should appear less frequently in older individuals than in younger ones, because carriers of the variant are slightly less likely to survive to old age or to have successfully reproduced.
After controlling for demographic factors that could produce spurious age-related frequency shifts, the team identified 168 variants showing this pattern. In 159 of the 168 cases, the variant was becoming rarer in older age brackets, consistent with purifying selection actively removing it from the population. In the remaining 9 cases, the variant was becoming more common with age, a signature of positive selection favoring it.
Ninety percent of the identified variants were extremely rare in the population. Seventy-one were already classified as pathogenic or likely pathogenic in ClinVar, the National Institutes of Health database that catalogues genetic variants known to cause disease in humans. Others were associated with cancer, cardiovascular disease, neurological conditions, kidney disease, and other serious health problems.
Human evolution is ongoing. It is happening right now, in living people, and it is detectable with modern genomic tools.
What purifying selection actually means
Purifying selection is the evolutionary process by which harmful genetic variants are gradually removed from a population over generations. It is the flip side of positive selection, in which beneficial variants spread because they increase survival or reproduction.
For purifying selection to leave a detectable signature in a living population, a variant needs to be harmful enough that its carriers are somewhat less likely to survive to older ages, either because the variant causes early death, reduces reproductive success, or both. The signal in the data is a shift in frequency across age groups: younger people carry the variant at a slightly higher rate than older people, because some carriers died or failed to reproduce before reaching old age.
The fact that this signal is detectable at all in a single biobank cohort of living adults, rather than requiring comparison of ancient DNA with modern DNA across thousands of years, is itself a methodological advance. Previous studies of natural selection in humans typically relied on either ancient DNA from archaeological remains or very long time-series comparisons of population genetics. This study demonstrates that a large cross-sectional biobank, comparing gene frequencies across age brackets within a single living population, can detect selection signals that would otherwise require far more elaborate historical data.
“Our findings reveal that purifying selection continues to act on rare, disease-linked variants in modern humans,” the researchers wrote. “This has implications for understanding the genetic architecture of complex diseases and for prioritizing rare variants in medical research.”
The specific variants being eliminated
The 168 variants identified by the study are not randomly distributed across the genome. They cluster in biologically meaningful locations, which adds confidence that the signals reflect genuine selection rather than statistical noise.
Among the most medically significant findings is the evidence of purifying selection acting against pathogenic variants in BRCA1, the gene most strongly associated with hereditary breast and ovarian cancer. Carriers of high-risk BRCA1 variants face substantially elevated lifetime risks of breast and ovarian cancer, and the study found evidence that these variants are being gradually removed from the Taiwanese population through natural selection. Carriers may be less likely to survive to old age or, in some cases, less likely to have had children, reducing the frequency of the variant in subsequent generations.
Alongside the BRCA1 findings, the study detected positive selection in nearby DNA repair regions, a pattern the researchers describe as complex pleiotropic dynamics. A single gene can affect multiple traits simultaneously, and the evolutionary forces acting on a gene may reflect a balance between its harmful effects in one context and its beneficial effects in another.
Other variants under purifying selection were linked to cardiovascular disease, neurological conditions including familial forms of Parkinson’s disease, kidney disease, and metabolic disorders. Several were previously categorized in medical databases as causing rare inherited conditions with serious health consequences.
The study also found evidence of positive selection acting on variants in two genes with particularly broad biological effects. ATG9A, which is involved in autophagy, the cellular process by which cells break down and recycle damaged components, showed a complex pattern of selection suggesting that variation in this gene is subject to competing evolutionary pressures. FADS2, which regulates fatty acid metabolism and has previously been associated with adaptation to different dietary environments, showed frequency shifts consistent with ongoing selection, possibly related to historical differences in diet across the populations that contributed to modern Taiwanese ancestry.
Why the variants are so rare
One of the study’s most significant methodological contributions is its focus on rare genetic variants, those present in fewer than one in 2,000 people in the sample. Previous genome-wide studies of natural selection in humans focused primarily on common variants, those present in at least one percent of the population, because rare variants are statistically harder to analyze. The Taiwan Biobank’s large sample size, over 72,000 individuals, provided enough statistical power to detect frequency shifts in variants that appear in only a handful of people per thousand.
This matters because the rarest variants tend to be the most recently arisen and the most harmful. Natural selection acts most efficiently against variants that cause serious, early-onset disease, and such variants tend to be rare precisely because selection has already been reducing their frequency over many generations. By focusing on this rare variant landscape, the study is detecting selection signals that most prior research missed.
The implication for medicine is direct. Variants that are being actively removed from the population by natural selection are, almost by definition, variants that cause significant harm to the people who carry them. The study’s list of 168 selected variants is therefore, in effect, a list of genetic variants that natural selection has identified as medically significant, not through human curation of clinical databases, but through the independent process of differential survival and reproduction playing out across generations.
“Evolutionary analyses can be valuable in prioritizing mutations that cause disease for future research,” the researchers wrote. The variants that natural selection is currently eliminating are the ones most likely to be worth studying as therapeutic targets.
What this does not mean
The finding that natural selection is detectable in living humans does not mean that the pace of human evolution is accelerating, or that modern medicine is failing to protect carriers of disease-causing variants. Natural selection has always operated on the rare end of the genetic spectrum, quietly reducing the frequency of the most harmful variants across many generations. What is new is not that selection is happening but that we now have genomic datasets large enough to detect it within a single living population rather than requiring ancient DNA comparisons.
The study is observational and correlational. The researchers cannot directly observe reproduction and survival in the way a controlled experiment would allow. The age-related frequency shifts they detected are consistent with purifying selection, but other explanations, including subtle population structure effects that were not fully controlled for, cannot be entirely ruled out.
The sample is drawn entirely from Han Taiwanese individuals enrolled in a voluntary biobank. Whether the same variants show the same selection signals in other populations with different genetic backgrounds, different disease environments, and different historical selection pressures is a question the study cannot answer. The researchers suggest that the analytical framework they developed could be applied to other large biobanks worldwide to test whether the signals they detected are population-specific or more universal.
The study also focused on variants with large individual effects on disease risk. The genetic variants that contribute most to common conditions like type 2 diabetes, hypertension, and depression tend to be common variants with small individual effects, and those fall below the detection threshold of this method. The selection signals captured here represent the visible tip of a much more complex evolutionary process.
What the study establishes, with unusual directness, is that the claim that human evolution has stopped is empirically wrong, at least at the level of rare, highly harmful genetic variants. Among 72,635 living people, the ancient process of natural selection is quietly, measurably, still at work.
The study, “Allele frequency trajectories across age groups reveal ongoing natural selection shaping disease susceptibility”, was authored by Jun-Lin Chen, Chia-Lin Hsieh, Pei-Lung Chen, James C. Liao, and Chun-Houh Chen at National Taiwan University and collaborating institutions, and published in the American Journal of Human Genetics in August 2026.
Source: National Taiwan University. DOI: 10.1016/j.ajhg.2026.07.002
