Abstract
Abstract
Rapid urbanization has reshaped human population structure, yet its impact on genetic architecture and disease risk remains poorly understood. Here, we present the
S
hanghai
H
uman
AN
cestry and
G
enomic
H
ealth
A
tlas
I
nitiative (SHANGHAI), a genomic resource comprising 1,016 individuals with high-coverage whole-genome sequencing (48.6×) integrated with multi-omics and clinical data, representing a rapidly urbanizing megacity population. We identified 43.2 million genetic variants, including 7.08 million previously unreported variants and over 100,000 structural variants, with substantial enrichment of rare and population-specific variation within this rapidly urbanizing megacity population. Despite sampling within a single metropolitan region, SHANGHAI exhibits pronounced fine-scale population structure driven by diverse ancestral origins, reflecting extensive migration and admixture. This hidden heterogeneity influences genetic association signals, reduces the portability of polygenic risk scores (PRS), and shapes the distribution of clinically relevant variants. Signals of recent positive selection differ across ancestry-defined subgroups and are enriched in pathways related to lipid metabolism and cardiovascular function, linking historical adaptation to present-day disease susceptibility. Notably, lineage-differentiated adaptive signals were detected at the
FADS
gene cluster and the
DOCK9
/
DOCK9-DT
locus, both significantly associated with lipid-related phenotypes. Functionally, rare and structural variants contribute disproportionately to high-impact genomic effects and extreme phenotypes. Integrating polygenic and rare variant effects, we demonstrate a layered genetic architecture for low-density lipoprotein cholesterol, in which polygenic background defines baseline risk while rare variants drive phenotypic extremes. Incorporating rare variants significantly improves disease risk prediction beyond PRS alone. Collectively, our findings illustrate that genetic heterogeneity within megacity populations has important implications for evolutionary inference and precision medicine. This study provides a framework for understanding genomic architecture in rapidly urbanizing populations and highlights the need for population-specific models in disease risk prediction.