Table 3.
Predictive value of the final models: area under the curve and discrimination parameters of the models obtained by the three linear discriminant (LD) binary classifiers.
| BD vs. CT | SZ vs. CT | BD vs. SZ | |
|---|---|---|---|
| Single domain—immune blood | |||
| General AUC | 0.73 | 0.71 | 0.75 |
| Accuracy | 69.17% | 70.30% | 71.73% |
| True positive rate (Sensitivity) | 72.31% | 71.48% | 79.10% |
| True negative rate (Specificity) | 64.39% | 61.27% | 64.49% |
| False positive rate (1—Specificity) | 35.61% | 38.73% | 35.51% |
| False negative rate (1—Sensitivity) | 27.69% | 28.52% | 20.90% |
| Single domain—cognition | |||
| General AUC | 0.81 | 0.90 | 0.77 |
| Accuracy | 75.35% | 87.06% | 72.14% |
| True positive rate (Sensitivity) | 76.12% | 84.43% | 71.39% |
| True negative rate (Specificity) | 71.19% | 81.14% | 71.27% |
| False positive rate (1—Specificity) | 28.81% | 18.86% | 28.73% |
| False negative rate (1—Sensitivity) | 23.88% | 15.53% | 28.61% |
| Multi-domain—immune blood + cognition | |||
| General AUC | 0.86 | 0.89 | 0.80 |
| Accuracy | 79.73% | 86.18% | 76.43% |
| True positive rate (Sensitivity) | 88.29% | 84.46% | 71.29% |
| True negative rate (Specificity) | 71.11% | 81.39% | 73.33% |
| False positive rate (1—Specificity) | 28.89% | 18.61% | 26.67% |
| False negative rate (1—Sensitivity) | 11.71.% | 15.54% | 28.71% |
AUC sensitivity (true positive rate), specificity (true negative rate), and false positive (1—specificity) and false negative (1—sensitivity) rates. Receiver operating characteristic (ROC) plots were derived from linear discriminative analysis (LDA) based on the top six biomarkers from the PLS-DA approach. Three different pairs of classifiers (BD vs. control, SZ vs. control, and BD vs. SZ) were analyzed using selected biomarkers from blood, cognition, and their combined pool. Optimal sensitivity and specificity of the set of selected biomarkers for the three models and were determined by the ROC curve and the AUC. An AUC of 0.90–1.0 is very high (excellent), of 0.80–0.89 high (good), of 0.70–0.79 moderate (fair), of 0.60–0.69 low (poor), and of 0.5–0.59 is very low (fail or useless). Color code: “success” showed in green, moderate “success” showed in yellow, “failure” showed in red.
AUC area under the curve, BD bipolar disorder, SZ schizophrenia, CT control.