For each of the five diseases, the score with the best discriminative capacity was determined based on maximal area under the receiver-operator curve (AUC) in a logistic regression model with the disease as the outcome and the disease-specific candidate GPS, age, sex, first four principal components of ancestry, and an indicator variable for genotyping array used (Supplementary Tables 1–6). AUC confidence intervals were calculated using the ‘pROC’ package within R.
