Table 4.
Sensitivity and specificity for all scores in the training and test data
| SNP selection | Score | Sens train (95 % CI) | Spec train (95 % CI) | Sens test (95 % CI) | Spec test (95 % CI) |
|---|---|---|---|---|---|
| 0.012 % | Allele count | 0.8256 (0.7890; 0.8571) | 0.8255 (0.7892; 0.8566) | 0.7532 (0.6938; 0.8044) | 0.8167 (0.7642; 0.8597) |
| (63 SNPs) | LogOR | 0.8025 (0.7644; 0.8358) | 0.8029 (0.7652; 0.8358) | 0.7489 (0.6892; 0.8005) | 0.8247 (0.7729; 0.8667) |
| LogReg | 0.8655 (0.8320; 0.8933) | 0.8645 (0.8312; 0.8920) | 0.7489 (0.6892; 0.8005) | 0.7928 (0.7385; 0.8384) | |
| Lasso | 0.8676 (0.8342; 0.8952) | 0.8686 (0.8357; 0.8957) | 0.7403 (0.6801; 0.7925) | 0.8008 (0.7470; 0.8455) | |
| RJ-Class | 1.0000 (0.9920; 1.0000) | 1.0000 (0.9922; 1.0000) | 0.7706 (0.7122; 0.8201) | 0.8207 (0.7685; 0.8632) | |
| 0.025 % | Allele count | 0.8130 (0.7755; 0.8455) | 0.8070 (0.7696; 0.8396) | 0.7489 (0.6892; 0.8005) | 0.8088 (0.7556; 0.8526) |
| (125 SNPs) | LogOR | 0.7689 (0.7290; 0.8045) | 0.7700 (0.7306; 0.8052) | 0.7143 (0.6529; 0.7687) | 0.7610 (0.7045; 0.8095) |
| Lasso | 0.8866 (0.8549; 0.9120) | 0.8871 (0.8559; 0.9122) | 0.7576 (0.6984; 0.8083) | 0.8088 (0.7556; 0.8526) | |
| RJ-Class | 1.0000 (0.9920; 1.0000) | 1.0000 (0.9922; 1.0000) | 0.7662 (0.7076; 0.8162) | 0.8207 (0.7685; 0.8632) | |
| 0.049 % | Allele count | 0.8529 (0.8183; 0.8819) | 0.8583 (0.8245; 0.8865) | 0.7532 (0.6938; 0.8044) | 0.7968 (0.7427; 0.8419) |
| (249 SNPs) | LogOR | 0.7773 (0.7378; 0.8124) | 0.7782 (0.7392; 0.8129) | 0.7273 (0.6665; 0.7806) | 0.7610 (0.7045; 0.8095) |
| Lasso | 0.9328 (0.9066; 0.9520) | 0.9322 (0.9064; 0.9513) | 0.7532 (0.6938; 0.8044) | 0.7968 (0.7427; 0.8419) | |
| RJ-Class | 1.0000 (0.9920; 1.0000) | 1.0000 (0.9922; 1.0000) | 0.7922 (0.7353; 0.8395) | 0.8088 (0.7556; 0.8526) | |
| 0.098 % | Allele count | 0.8782 (0.8457; 0.9045) | 0.8665 (0.8334; 0.8939) | 0.7359 (0.6756; 0.7886) | 0.8207 (0.7685; 0.8632) |
| (496 SNPs) | LogOR | 0.7983 (0.7599; 0.8319) | 0.7967 (0.7587; 0.8301) | 0.7316 (0.6710; 0.7846) | 0.7649 (0.7087; 0.8132) |
| Lasso | 0.9622 (0.9410; 0.9759) | 0.9671 (0.9473; 0.9797) | 0.7056 (0.6439; 0.7607) | 0.8207 (0.7685; 0.8632) | |
| RJ-Class | 1.0000 (0.9920; 1.0000) | 1.0000 (0.9922; 1.0000) | 0.8009 (0.7446; 0.8473) | 0.8048 (0.7513; 0.8491) | |
| 0.196 % | Allele count | 0.9223 (0.8947; 0.9431) | 0.9138 (0.8855; 0.9356) | 0.7143 (0.6529; 0.7687) | 0.7849 (0.7299; 0.8312) |
| (991 SNPs) | LogOR | 0.8256 (0.7890; 0.8571) | 0.8255 (0.7892; 0.8566) | 0.7316 (0.6710; 0.7846) | 0.7849 (0.7299; 0.8312) |
| Lasso | 0.9790 (0.9618; 0.9885) | 0.9795 (0.9626; 0.9888) | 0.7056 (0.6439; 0.7607) | 0.8406 (0.7903; 0.8807) | |
| RJ-Class | 1.0000 (0.9920; 1.0000) | 1.0000 (0.9922; 1.0000) | 0.7965 (0.7400; 0.8434) | 0.7809 (0.7257; 0.8276) | |
| 0.782 % | Allele count | 0.9370 (0.9115; 0.9555) | 0.9363 (0.9111; 0.9548) | 0.6061 (0.5418; 0.6668) | 0.7092 (0.6502; 0.7619) |
| (3,960 SNPs) | LogOR | 0.8971 (0.8665; 0.9213) | 0.8973 (0.8672; 0.9213) | 0.7143 (0.6529; 0.7687) | 0.8127 (0.7599; 0.8562) |
| Lasso | 1.0000 (0.9920; 1.0000) | 1.0000 (0.9922; 1.0000) | 0.6926 (0.6304; 0.7486) | 0.8327 (0.7816; 0.8738) | |
| RJ-Class | 1.0000 (0.9920; 1.0000) | 1.0000 (0.9922; 1.0000) | 0.7792 (0.7214; 0.8279) | 0.7610 (0.7045; 0.8095) | |
| 3.125 % | Allele count | 0.9685 (0.9487; 0.9808) | 0.9671 (0.9473; 0.9797) | 0.5455 (0.4810; 6084) | 0.6175 (0.5561; 0.6755) |
| (15,835 SNPs) | LogOR | 0.9832 (0.9672; 0.9915) | 0.9836 (0.9679; 0.9917) | 0.7576 (0.6984; 0.8083) | 0.7689 (0.7130; 0.8168) |
| Lasso | 1.0000 (0.9920; 1.0000) | 1.0000 (0.9922; 1.0000) | 0.7792 (0.7214; 0.8279) | 0.7928 (0.7385; 0.8384) | |
| RJ-Class | 1.0000 (0.9920; 1.0000) | 1.0000 (0.9922; 1.0000) | 0.7532 (0.6938; 0.8044) | 0.7649 (0.7087; 0.8132) | |
| 12.5 % | LogOR | 1.0000 (0.9920; 1.0000) | 1.0000 (0.9922; 1.0000) | 0.6883 (0.6259; 0.7446) | 0.7490 (0.6919; 0.7986) |
| (63,334 SNPs) | RJ-Class | 1.0000 (0.9920; 1.0000) | 1.0000 (0.9922; 1.0000) | 0.7446 (0.6847; 0.7965) | 0.7769 (0.7214; 0.8240) |
Sens sensitivity, CI confidence interval, spec specificity, Lasso least absolute shrinkage and selection operator, RJ-Class Random Jungle classification
Allele count: score constructed based on number of risk alleles
Log OR: score constructed by weighting variants with respective log odds ratio from single marker analyses
LogReg: score constructed from logistic regression