The largest ancient DNA dataset assembled from 15,836 western Eurasian individuals identified 479 gene variants under directional selection after the transition to agriculture, with changes accelerating in the Bronze Age around 5,000 years ago (Nature, 2025). Immunity-linked variants were most frequent targets; a tuberculosis-susceptibility allele declined over the past 3,000 years while a multiple-sclerosis risk variant rose 6,000 years ago before receding in some groups in the last 2,000 years (Nature, 2025; Cell, 2022).

Reich and Akbari's method isolated consistent frequency shifts across time slices to separate selection from drift and migrations such as the influx of Anatolian farmers (Nature, 2025). This expands on the canonical lactose-persistence allele to hundreds of loci affecting metabolism, immunity and complex traits (Nature Genetics, 2022).

AI-driven variant-effect models and large-scale GWAS enabled scaling the analysis beyond prior studies that detected only sparse signals; the same computational frameworks now underpin polygenic-risk scoring in biotech and direct detection of recent selection in UK Biobank cohorts (Nature Machine Intelligence, 2023; Nature, 2025). Original coverage omitted the explicit pipeline reliance on machine-learning denoising of 10,000 newly sequenced genomes and downstream implications for CRISPR targeting of post-Neolithic alleles.