Inferring the demographic history of Hexi Corridor over the past two millennia from ancient genomes

The Hexi Corridor in northwest China has for millennia served as a crossroads for interactions between East Asia, Central Asia and lands further west, while also connecting the Chinese heartland and the northern Mongolian Steppe and Tibetan Plateau (TP) to the south (Fig. 1a). Previous studies have revealed the details of this cross-continental material and cultural transmission as early as the Bronze Age onwards. Materials and technologies exchanged along the ancient Silk Road included East Asian millets and painted pottery, West Asian wheat and barley, bronze metallurgy, domesticated sheep, horses, and cattle [1]. To date, paleogenomic studies have examined the genetic structure of the eastern Hexi Corridor and neighboring populations. To address the limitations above, we sampled and sequenced 30 ancient individuals across the Hexi Corridor (Fig. 1a,b and Table S1A online). We sampled 17 human individuals from the Heishuiguo site from the central Hexi Corridor, dating from the Western Han Dynasty (202 BCE–8 CE) through the Eastern Han Dynasty (25–220 CE). From the western Hexi Corridor, we sampled 13 human individuals from the Foyemiaowan cemetery, dating from the Cao-Wei (220–265 CE) to the Tang dynasties (618–907 CE). To establish the chronology of our study, we relied on both archaeological and radiocarbon dating evidence. We list the detailed information of each sample in Table S1A (online). We constructed double-stranded DNA libraries from teeth or petrous portions of the temporal bone for the studied individuals. We generated shotgun sequencing data that revealed high levels of DNA preservation with endogenous DNA contents ranging from 0.95% to 83.60% (28.79% on average) (Table S1A online). All samples showed characteristics of postmortem damage patterns (Fig. S1 online). The pseudo-haploid genotype was called on the 1.24 million SNP panel, resulting in 14,387 to 1,141,519 SNP sites. Of the 30 Hexi Corridor individuals sequenced, we detected one second-degree kinship confirmed by three methods, and we removed relatives with fewer SNP counts. Finally, we used 25 individuals with