Table 3.

Calculation of an Optimal ROC Curve: Adding Five SNPs—One Shown—to the Calculations in Table 2

Environmental and Genetic Risk Factors Combination Frequencies from Previous Testa (Case;Control)	Genotype Freq.b (Case;Control)		Genotype Freq. (Case;Population)		Environmental and Genetic Risk Factors Combination Frequencies (Case;Control)	Likelihood Ratio	Rank Percentagec
Dietary Score x … x rs5219x …	rs4402960	…	rs4402960	…	Dietary Score x … x rs4402960 x …
8.238e-09;1.450e-09	0.103;0.115	…	0.103;0.114	…	1.118e-11;1.160e-12	9.640	0.02%
…	…	…	…	…	…	…	…
1.275e-08;2.767e-09	0.103;0.115	…	0.103;0.114	…	1.731e-11;2.300e-12	7.523	0.05%
…	…	…	…	…	…	…	…
7.189e-09;1.806e-09	0.103;0.115	…	0.103;0.114	…	9.756e-12;1.534e-12	6.359	0.10%
…	…	…	…	…	…	…	…
9.568e-09;2.869e-09	0.103;0.115	…	0.103;0.114	…	1.298e-11;2.492e-12	5.210	0.21%
…	…	…	…	…	…	…	…
4.037e-09;1.767e-09	0.103;0.115	…	0.103;0.114	…	5.478e-12;1.589e-12	3.447	0.89%
…	…	…	…	…	…	…	…

Details of using the optimal ROC-curve approach to further improve the new predictive genetic test in Table 2 by incorporating five additional novel risk SNPs discovered in the second genome-wide association study of type 2 diabetes¹⁰ (five combinations only, out of a total of $2\times {10}^8$ combinations).

^aMultilocus genotype probabilities given disease status obtained from the new predictive genetic test in Table 2.

^bGenotype frequency estimates obtained from the confirmatory stage (stage 2) of the second genome-wide association study.¹⁰

^c100•(rank of LR)/(total number of combinations).