Traces of directional selection at hundreds of loci in DNA from 15,836 ancient West Eurasians
Background: Distinguishing migration from natural selection
An analysis of DNA from 15,836 ancient West Eurasians found evidence that directional selection acted at hundreds of genomic loci over the past 10,000 years. Published in Nature, the study’s central contribution is an analytical method for distinguishing persistent natural selection from changes caused by migration and admixture.
An allele is one of the different genetic forms found at the same genomic position. Directional selection occurs when an allele that benefits its carriers’ survival or reproduction rises persistently in frequency over multiple generations. A change in frequency alone, however, cannot establish selection: migration, admixture and random genetic drift can also alter frequencies.
The phenomenon that earlier studies reported to be rare across human evolution was the classic selective sweep, in which a new advantageous variant becomes fixed in a population. This analysis also examined sustained frequency changes that had not reached fixation.
Key findings: Strong selection signals at 479 loci
The ancient samples span 18,000 years, and new data for 10,016 individuals were reported in this paper. The analysis also included 6,438 modern individuals. The full time span of the ancient samples should be distinguished from the past 10,000 years, the period over which the principal selection results were interpreted.
After accounting for population structure using genetic similarity among individuals, the researchers tested whether sample age additionally explained a consistent trend in allele frequencies. They calculated selection coefficients across approximately 9.7 million variants after statistically imputing missing genotypes and applying quality control. The single-variant tests used 13,936 ancient individuals, excluding close relatives.
The analysis identified 479 independent loci with a posterior probability of selection greater than 99%. Excluding the human leukocyte antigen (HLA) region, which is dense with immune-related genes, left 410 loci. For variants affecting blood, immune, inflammatory, cardiovascular or metabolic traits, selection intensity increased in the Bronze Age compared with the pre-agricultural period. The researchers proposed dietary change, population growth and close contact with livestock as possible explanations, but did not directly establish a causal link to any of them.
Nine loci associated with lighter skin pigmentation passed the genome-wide significance threshold, and selection was suggested at one additional locus. Adaptation to low sunlight and dietary vitamin D deficiency remains the researchers’ interpretation. At some loci, including TYK2, traces of reversals in the direction of selection over time were also observed.
Meaning and outlook: Widespread selection and a limited overall contribution
Finding many selection signals does not mean that directional selection explains most genomic change. In the researchers’ model, directional selection accounted for an average of 2.16% of the variance in allele frequencies on the logit scale—that is, after frequencies are transformed to log odds. The methods section reports a jackknife standard error of 0.11 percentage points. Together, these figures show both widespread selection signals and a limited overall contribution.
Trait-level interpretations also call for caution. It is not established whether variants associated with disease or physical traits in modern genome-wide association studies (GWAS) had the same effects in ancient populations, or whether those traits were themselves the targets of selection. A single variant may also affect multiple traits.
The primary detection model assumes constant selection pressure over time and across space, so it may miss cases in which the direction of selection changed. A task for follow-up studies applying the same approach to other regions and longer time series is to distinguish patterns specific to West Eurasia’s recent history from those shared more broadly. The results and methods are available in the original Nature paper: https://www.nature.com/articles/s41586-026-10358-1
Nature, Published online: 15 April 2026; doi:10.1038/s41586-026-10358-1Analysis of 15,836 ancient West Eurasian genomes reveals hundreds of instances of directional selection, showing that sustained changes in allele frequency were widespread, rather than being rare over this period as previously assumed.
This study provides a foundation for examining how the frequencies of modern disease-associated variants changed in the past. For example, a team studying immune-disease loci could compare the selection estimates and frequency trajectories for approximately 9.7 million variants with modern GWAS findings to narrow down candidates for follow-up functional experiments. The timing of a variant’s increase or decrease, and reversals in selection direction, provide starting points for hypotheses testing relationships with past environments. However, a marker representing a selection signal is not necessarily the causal variant that was actually targeted by selection. After selecting research candidates, researchers still need to test the relationships between function and environment separately. These data alone cannot determine an individual’s disease risk or the effects of treatment or public-health policy.