Genome-wide associations of structural variants with human traits through imputation from long-read assemblies

Structural variants (SVs) are a major type of genetic variation, yet their role in human traits remains largely uncharacterized, primarily due to challenges in genotyping them on a genome-wide scale in large cohorts. Here we identified 171,233 high-quality, genome-wide SVs from 482 haplotype-resolved genome assemblies derived from PacBio HiFi long-read sequencing of 241 individuals. We developed a reference panel and a web application (ImputeSV) to impute these SVs from single-nucleotide polymorphism (SNP) data and demonstrated high imputation accuracy at both the individual and cohort levels. Using this tool, we imputed 54,578 common SVs (minor allele frequencies (MAFs) ≥1%) in 456,643 UK Biobank (UKB) participants of European ancestry. Through analysis of UKB data and simulations, we estimated that SVs contributed to at least 4.7% of the common genetic variation for complex traits. Genome-wide association analyses of SVs for 2,624 UKB traits identified 17,335 SV-trait associations, including 958 unlikely to be driven by small genetic variants. Our study demonstrates the power of using long-read assemblies for imputing SVs from SNPs, unveils the role of SVs in complex trait variation and provides a catalog of SV associations in the UKB.