Abstract
The extent to which human adaptations have persisted throughout history despite strong eroding demographic events such as admixture, genetic drift, and fluctuations in selection pressures remains unknown. Understanding which loci are particularly resilient to such forces may shed light on the traits that were important for humans throughout multiple time periods. Yet, detecting ancient selection events is challenging from modern and ancient DNA due to the data and/or signal being severely degraded. Here we use a domain-adaptive neural network (DANN) trained on simulated data and applied to ancient and modern DNA for sweep detection. We show that the DANN can account for simulation misspecification, or discrepancies between the simulations and real aDNA, thereby improving the ability to detect sweeps in real data. Application of the DANN to more than 800 ancient and modern human genomes spanning the last 7000 years recovered 16 known sweeps at loci including LCT, HLA, KITLG, and OCA2/HERC2, and revealed 32 novel sweeps. All identified sweeps were classified as hard, consistent with historically low population sizes. While some sweeps were lost over time, 14 sweeps at loci involved in a range of functions including neuronal, reproductive, pigmentation, and signaling traits were found to persist from the most ancient time periods into the most recent time periods. Notably, the same top haplotype remained at high frequency across time at 9 of these 14 sweeps. Together, these results indicate that hard sweeps predominated in ancient human history and that several ancient selective events were resilient to strong admixture events and experienced sustained selective pressures.