Genotype imputation is essential for enhancing the power of association-mapping and discovering rare and indels that are missed by most genotyping arrays. Imputation analysis can be more accurate with a population-specific reference panel or a multi-ethnic reference panel with numerous samples. The National Institute of Health, Republic of Korea, initiated the Korean Reference Genome (KRG) project to identify variants in whole-genome sequences of & SIM;20,000 Korean participants. In the pilot phase, we analyzed the data from 1,490 participants. The genetic characteristics and imputation performance of the KRG were compared with those of the 1,000 Genomes Project Phase 3, GenomeAsia 100K Project, ChinaMAP, NARD, and TOPMed reference panels. For comparison analysis, genotype panels were artificially generated using whole-genome sequencing data from combinations of four different ancestries (Korean, Japanese, Chinese, and European) and two population-specific optimized microarrays (Korea Biobank Array and UK Biobank Array). The KRG reference panel performed best for the Korean population (R (2) = 0.78-0.84, percentage of well-imputed is 91.9% for allele frequency > 5%), although the other reference panels comprised a larger number of samples with genetically different background. By comparing multiple reference panels and multi-ethnic genotype panels, optimal imputation was obtained using reference panels from genetically related populations and a population-optimized microarray. Indeed, the reference panels of KRG and TOPMed showed the best performance when applied to the genotype panels of KBA (R (2) = 0.84) and UKB (R (2) = 0.87), respectively. Using a meta-imputation approach to merge imputation results from different reference panels increased the imputation accuracy for rare variants (& SIM;7%) and provided additional well-imputed variants (& SIM;20%) with comparable imputation accuracy to that of the KRG. Our results demonstrate the importance of using a population-specific reference panel and meta-imputation to assess a substantial number of accurately imputed rare variants.
