Figure 2: Risk score prediction of disease is most accurate when including association data from multiple racial groups.
A. Single-population association risk scores were assessed using logistic regression models, and the significance of the risk score as a predictor of IBD state is shown. B. AUC values calculated using logistic regression models which predicted IBD status using age, sex, and the combination of predictors on the x-axis. Cross-validation was also performed to test for overfitting. Association data used for risk scores: Liu included European individuals, Brant included African American individuals, and Hui included AJ individuals.