Table 1.
Parameters and notation of polygenic model.
Figure 1.
Expected −log10(P) of linear regression estimate as a function of P-value threshold for selecting markers into the polygenic score.
Training sample, 3322 cases and 3587 controls; replication sample, 2687 cases and 2656 controls. Marker panel of 74062 independent SNPs. Variance explained by markers, 28.7%. pi0, proportion of markers with no effect on disease.
Figure 2.
Expected −log10(P) of allele score estimate as a function of P-value threshold for selecting markers into the polygenic score.
Training sample, 3322 cases and 3587 controls; replication sample, 2687 cases and 2656 controls. Marker panel of 74062 independent SNPs. Variance explained by markers, 28.7%. pi0, proportion of markers with no effect on disease.
Table 2.
AUC calculated by Evans et al [19] compared to analytic values when () marker panel explains half the heritability, or (
) marker panel explains the full heritability.
Table 3.
R2 reported for complex diseases compared to analytic values when marker panel explains one quarter, one half or the full heritability.
Figure 3.
AUC as a function of sample size, using a panel of 100,000 markers that explains half the heritability of liability.
n, number of cases and of controls in training sample. Heritability of liability, 76% for Crohn's disease. 44% for breast cancer. Line annotations are the proportion of markers with no effect on disease.
Table 4.
Numbers of cases and controls (in 1000s of each, rounded up) required to attain a specified AUC using a panel of 100,000 markers that explains half the heritability of liability.
Figure 4.
AUC as a function of sample size, using a panel of 1,000,000 markers that explains the full heritability.
n, number of cases and of controls in training sample. Heritability of liability, 76% for Crohn's disease. 44% for breast cancer. Line annotations are the proportion of markers with no effect on disease.
Table 5.
Numbers of cases and controls (in 1000s of each, rounded up) required to attain a specified AUC using a panel of 1,000,000 markers that explains the full heritability.
Table 6.
Numbers of subjects (in 1000s, rounded up) required to attain a specified correlation with a normal trait using a panel of 1,000,000 markers that explains the full heritability.