Fig. 3

Imputation accuracy of lpWGS using high-coverage WES variants as ground truth. a The number of variants both imputed by lpWGS and called by high-coverage WES per sample and the proportion of these variants with imputation quality score (INFO) at least 0.8. b Boxen plot of the dosage correlation per sample grouped by imputation quality score intervals. Dosage correlation is defined as the correlation between the imputed variant dosage, which is a continuous value ranging from 0 to 2, and the number of alternate alleles called from WES. c The dosage correlation of lpWGS imputed variants grouped by cohort minor allele frequency before and after filtering out variants with imputation quality score below 0.8. We observe consistently strong performance after filtering regardless of MAF bins. d The Pearson correlation between the allele frequency of the imputed variants in our cohort versus their allele frequencies in gnomAD ancestry groups