Imagine thousands of people gathered for an enormous birthday party. The 40-year-olds stand in one corner, the 60-year-olds in another, and those over 70 at the far end of the room.
Now imagine being able to look inside their DNA.
DNA is like an instruction manual written with billions of letters. The text is nearly identical in all of us, but some letters differ. Scientists call these differences genetic variants.
When researchers compared people of different ages, they noticed something unusual.
A variant associated with Alzheimer's disease was more common among younger people than among those over 70.
It had not disappeared from anyone's DNA. Researchers simply found it less often among people who had lived beyond 70.
That suggested the variant might be connected to lifespan.
But why?
Researchers investigated by arranging the genetic data of tens of thousands of people by age – recreating our imaginary party on a computer.
Their reasoning was simple: If a genetic feature reduces people's chances of reaching old age, it should appear less often among older groups.
In a large-scale study, the team examined genetic data from 57,696 people of European ancestry in the Genetic Epidemiology Research on Adult Health and Aging cohort.
Among millions of genetic differences, one produced an especially strong signal: APOE ε4, which is known to increase the risk of late-onset Alzheimer's disease and cardiovascular disease.
The proportion of people carrying it began to decline after roughly age 70.
"What we directly showed is that APOE ε4 is associated with survival," geneticist Hakhamanesh Mostafavi, an assistant professor at New York University Grossman School of Medicine, told ScienceAlert.
"Its frequency declines among individuals beyond about age 70."
Carrying APOE ε4 does not mean someone will inevitably develop Alzheimer's. It is a risk factor, not a diagnosis.
Still, the finding raises an intriguing question.
Alzheimer's usually develops after people have had children. So why would natural selection act against a genetic feature associated with a late-life disease?
Survival is only one part of evolution. For a variant to become less common across generations, it must ultimately reduce its chances of being passed on.
Men can father children later in life, so survival could still affect reproduction. The variant might also influence health or fertility long before Alzheimer's appears.
Genes, after all, rarely do only one job.
"So we should not conclude that these variants have been selected against specifically because they cause late-life disease," Mostafavi cautions.
The researchers found a second signal. This time, it involved smoking.
The team examined genetic data from more than 117,000 UK Biobank participants. The participants had also reported how long their parents lived.
The parents had not been tested. But children inherit DNA from their parents, so the participants' genes offered indirect clues.
One variant near CHRNA3 is associated with heavier smoking. Researchers found it more often in people whose fathers had died younger.
In short, the smoking-related variant was linked to shorter paternal lifespans.
The pattern was not clear among mothers, perhaps because men in those generations smoked more heavily.
This did not prove the variant caused their deaths. Researchers lacked the fathers' DNA and smoking histories. But the association showed how genes and behavior may work together.
Yet the most surprising result may have been what the researchers did not find.
The study was powerful enough to detect many common variants with effects as large as the Alzheimer's-linked one. But only two major individual signals stood out: APOE and CHRNA3.
Where were all the others?
"This indirectly suggests that variants with large detrimental effects even late in life may generally not reach high frequencies in the population," Mostafavi says.
Their absence may itself be a mark of evolution. Natural selection may have prevented many harmful variants from becoming common in the modern human population.
Nearly a decade later, larger datasets can reveal rarer variants. The APOE result has also been replicated, Mostafavi says.
But the central picture remains: longevity reflects many small genetic influences, along with behavior, environment, healthcare, and chance.
The study cannot show why fewer people over 70 carried APOE ε4. Alzheimer's, cardiovascular disease, or another effect may have contributed.
Nor does it prove natural selection is rapidly removing the variant. Its effects on reproduction remain unclear.
Related: Humans Are Still Evolving Before Our Eyes on The Tibetan Plateau
The findings join other evidence that humans have not stopped evolving, from adaptations to high-altitude life on the Tibetan Plateau to the increasingly common median artery in the human forearm.
Evolution may still be moving quietly through our imaginary birthday party – one year and one empty chair at a time.
The study was published in PLOS Biology.