We performed a linear regression using PRSice-2, with the UK Biobank data as target sample using the default parameters. When PRS is calculated from the best-guess genotype, the best-guess genotype is defined as the genotype having an imputation probability of ≥0.9. If there is no such genotype, then the SNP is considered to be missing for the individual. In addition, for the imputed data, we filtered out SNPs with imputation quality score <0.8. With height as the outcome and PRS for height as predictor, we observed an increase in phenotypic variance explained (R2) of the PRS from 0.145 when using genotyped data to 0.152 when using best-guess imputed genotypes, and 0.153 when using dosage data; likewise, the R2 for BMI increased from 0.0475 when using genotype data to 0.0529 when using best-guess genotypes, and to 0.0535 when using dosage data. These results exemplify the potential gain in predictive power when using dosage data compared to using genotyped or best-guess genotype data. However, given the modest increases in predictive power, users may wish to perform first-pass analyses on genotyped-only data before
