351: When Selection Survives Admixture: Hard Sweeps in Ancient Eurasians

Harris M et al., Proceedings of the National Academy of Sciences (PNAS) - This episode examines a PNAS study that uses a domain-adaptive neural network to detect and classify selective sweeps in over 800 ancient and modern Eurasian genomes spanning ~7,000 years. The work recovers known targets (HLA, LCT, OCA2/HERC2, KITLG), reports 32 novel ancient sweep candidates, finds hard sweeps predominate, and shows 14 sweeps persisted across a major admixture event, highlighting resilience of certain adaptations. Key terms: ancient DNA, selective sweeps, domain-adaptive neural network, hard sweeps, admixture. Study Highlights:The authors trained a domain-adaptive neural network on simulated and ancient DNA to distinguish hard sweeps, soft sweeps, and neutrality across 708 ancient and 99 modern Eurasian genomes. The DANN outperformed standard CNNs under demographic and missing-data misspecification and detected 48 ancient sweeps, including 16 overlapping prior reports and 32 novel candidates. All identified sweeps were classified as hard and, after accounting for misclassification rates, the majority remain best explained by hard sweeps. Fourteen sweeps at genes involved in neuronal, reproductive, pigmentation, and signaling functions persisted across a major admixture event, often retaining the same high-frequency haplotype. Conclusion:Domain-adaptive deep learning improves detection of selective sweeps in degraded ancient genomes; hard sweeps were the dominant mode of adaptation in these ancient Eurasian samples and several selective events persisted despite strong admixture, pointing to sustained functional importance of particular loci. Music:Enjoy the music based on this article at the end of the episode. Article title:The persistence and loss of hard selective sweeps amid admixture in ancient Eurasians First author:Harris M Journal:Proceedings of the National Academy of Sciences (PNAS) DOI:10.1073/pnas.2528672123 Reference:Harris M., Mo Z., Siepel A., Garud N.R. The persistence and loss of hard selective sweeps amid admixture in ancient Eurasians. Proc. Natl. Acad. Sci. U.S.A. 2026;123(17):e2528672123. doi:10.1073/pnas.2528672123 License:This episode is based on an open-access article published under the Creative Commons Attribution 4.0 International License (CC BY 4.0) – https://creativecommons.org/licenses/by/4.0/ Support:Base by Base – Stripe donations: https://donate.stripe.com/7sY4gz71B2sN3RWac5gEg00 Official website https://basebybase.com On PaperCast Base by Base you'll discover the latest in genomics, functional genomics, structural genomics, and proteomics. Episode link: https://basebybase.com/episodes/hard-sweeps-admixture-ancient-eurasians QC:This episode was checked against the original article PDF and publication metadata for the episode release published on 2026-04-26. QC Scope:- article metadata and core scientific claims from the narration- excludes analogies, intro/outro, and music- transcript coverage: Audited the transcript sections describing the DANN method (domain adaptation, GRL, haplotype image inputs), the major results (counts of ancient/modern sweeps, hard sweeps, persistence across admixture), and the highlighted loci (HLA, LCT, OCA2/HERC2, KITLG), plus the discussion of admixture timing (~4.5 kya) and impl- transcript topics: Domain-adaptive neural networks (DANN) and gradient reversal layer; Ancient DNA data quality and missing data; Hard vs soft sweeps definitions; Sweep detection results (ancient vs modern) and total counts; Admixture event around 4.5 kya and sweep persistence; Loci of interest: HLA, LCT, OCA2/HERC2, KITLG QC Summary:- factual score: 10/10- metadata score: 10/10- supported core claims: 5- claims flagged for review: 0- metadata checks pass... Chapters (00:00:21) - How Your DNA Is a Survival Journal(00:01:20) - The Hidden Story of Human Evolution(00:07:37) - The AI Classifies Ancient Selection Swops(00:14:27) - How European genetic diversity survived the Bronze Age(00:18:42) - The Hidden History of Human Evolution