A haplotype reference panel constructed from 490,319 UK Biobank genomes improves genotype imputation for global populations

Here we constructed a genotype imputation reference panel from whole-genome sequencing data from 490,319 UK Biobank (UKB) participants¹ and enabled both standalone and cloud-based phasing and imputation through EagleImp² and EagleImp-RAP within the UK Biobank Research Analysis Platform. The UKB reference panel achieves lower phasing switch error rates than the widely used TOPMed r3 panel (133,597 individuals) for all global superpopulations except AMR and for 25 of 26 1000 Genomes Project subpopulations. Imputation accuracy, measured by MAE, improved for 4 of 5 superpopulations and 12 subpopulations for variants with minor allele frequencies down to 0.01%. Estimated squared correlation further improved across all 5 superpopulations and 22 subpopulations, with the largest gains in East and South Asian populations. Reimputation of COVID-19 genome-wide association study datasets from Italy, Germany, Spain and Norway^3,4 demonstrated that common-variant imputation remains unsaturated, enhancing the discovery of genome-wide significant loci and improving fine-mapping resolution.