Fig. 3: Best method for each disease and quantitative trait.
a The method that produced the highest odds ratio per standard deviation (OR/SD) in each disease was chosen as the best method. The median OR/SD of PGS with 95% confidence interval for 19 complex diseases was assessed in logistic regression models adjusted for age, sex, genotyping array, and the first ten principal components of ancestry in the testing group of N  =  188,745 independent UK Biobank unrelated European individuals. b PGS methods’ performance of frequency as the best method across diseases. Bars represent the times chosen as the best method across 19 diseases. c The method that produced the highest Pearson R2 in each quantitative trait was chosen as the best method. The median Pearson R2 with 95% confidence interval was calculated using Pearson’s correlation coefficient on the residuals from a linear regression model of the specific PGS with age, sex, array, and the first ten principal components, with a linear regression model of the particular quantitative trait outcome with age and sex. Evaluations were conducted in N  =  188,745 independent UK Biobank unrelated European testing individuals. d PGS methods’ performance of frequency as the best method across quantitative traits. Bars represent the times chosen as the best method across quantitative traits.
