Table 3.
Evaluation results of ML methods on (a) the full data set and (b) the suspected diagnosis data set. Methods are applied to all 53 features, 5 features selected with ANOVA, or LDA dimensions. The methods were evaluated by false negatives (FN) and false positives (FP) on the training and test set. Abbreviations: Asa—argininosuccinate, aas—age at sampling, BIO—biotinidase, C14OH—3-OH-tetradecanoylcarnitine, C16:1OH—3-OH hexadecenoylcarnitine, C5—isovalerylcarnitine, FN—false negatives, FP—false positives, His—histidine, LDA—linear discriminant analysis, LR—logistic regression, MeGlut—3-methylglutarylcarnitine, RR—ridge logistic regression, SVM—support vector machine, Trp—tryptophan, Val—valine.
| Method | Features (Number) | Train FN | Train FP | Test FN | Test FP |
|---|---|---|---|---|---|
| (a) Full data set | |||||
| LR | all (53) | 0 | 65 | 0 | 27 |
| RR | all(53) | 6 | 20,065 | 3 | 5005 |
| SVM | all(53) | 1 | 35 | 0 | 9 |
| LR | C5, C16:1OH, aas, Val, BIO | 0 | 167 | 0 | 42 |
| RR | C5, C16:1OH, aas, Val, BIO | 5 | 6026 | 1 | 1577 |
| SVM | C5, C16:1OH, aas, Val, BIO | 1 | 68 | 0 | 15 |
| (b) Suspected diagnosis data set | |||||
| LR | all (53) | 0 | 29 | 0 | 7 |
| RR | all (53) | 0 | 18 | 0 | 5 |
| SVM | all (53) | 0 | 20 | 0 | 6 |
| LR | Trp, C14OH, MeGlut, His, Asa | 0 | 35 | 0 | 6 |
| RR | Trp, C14OH, MeGlut, His, Asa | 0 | 35 | 0 | 6 |
| SVM | Trp, C14OH, MeGlut, His, Asa | 0 | 37 | 0 | 6 |
| LDA-LR | LDA dimensions | 0 | 9 | 0 | 10 |
| LDA-RR | LDA dimensions | 0 | 22 | 0 | 12 |
| LDA-SVM | LDA dimensions | 0 | 12 | 0 | 10 |