Table 2 Area under the ROC curve (AUROC) and seven statistical measures of random forest models for Hoehn & Yahr (HY) stages 1–3, 1, 2, and 3.
AUROC | 95% Confidence Interval | Accuracy | Positive Predictive Value (PPV) | Sensitivity | Specificity | F1 score | Negative Predictive Value (NPV) | MCC | |||
|---|---|---|---|---|---|---|---|---|---|---|---|
Nested cross-validation | Bacterial feature-based models | HY stages 1–3 | 0.548 | 0.456–0.641 | – | – | – | – | – | – | – |
HY stage 1 | 0.799 | 0.615–0.982 | 79.2% | 81.8% | 75.0% | 83.3% | 78.3% | 76.9% | 0.585 | ||
HY stage 2 | 0.705 | 0.592–0.818 | 69.4% | 54.8% | 58.6% | 75.0% | 56.7% | 77.8% | 0.331 | ||
HY stage 3 | 0.509 | 0.301–0.719 | – | – | – | – | – | – | – | ||
Clinical feature-based model | HY stages 1–3 | 0.559 | 0.464–0.654 | – | – | – | – | – | – | – | |
HY stage 1 | 0.549 | 0.307–0.791 | – | – | – | – | – | – | – | ||
HY stage 2 | 0.719 | 0.602–0.835 | 70.6% | 55.3% | 72.4% | 69.6% | 62.7% | 83.0% | 0.401 | ||
HY stage 3 | 0.772 | 0.619–0.925 | 66.1% | 40.0% | 92.3% | 58.1% | 55.8% | 96.2% | 0.427 | ||
Cross-validation | Bacterial feature-based models | HY stages 1–3 | 0.677 | 0.592–0.762 | 60.6% | 44.0% | 83.0% | 50.0% | 57.5% | 86.2% | 0.316 |
HY stage 1 | 0.868 | 0.719–1.000 | 83.3% | 83.3% | 83.3% | 83.3% | 83.3% | 83.3% | 0.667 | ||
HY stage 2 | 0.793 | 0.692–0.895 | 77.7% | 67.9% | 65.5% | 83.9% | 66.7% | 82.5% | 0.499 | ||
HY stage 3 | 0.705 | 0.544–0.865 | 67.9% | 36.4% | 66.7% | 68.2% | 47.0% | 88.2% | 0.293 | ||
Fusicatenibacter | HY stage 1 | 0.861 | 0.707–1.000 | 83.3% | 78.6% | 91.7% | 75.0% | 84.6% | 90.0% | 0.676 | |
Clinical feature-based model | HY stages 1–3 | 0.656 | 0.571–0.741 | 55.8% | 41.9% | 90.7% | 38.7% | 57.3% | 89.6% | 0.305 | |
HY stage 1 | 0.639 | 0.402–0.876 | 70.8% | 69.2% | 75.0% | 66.7% | 72.0% | 72.7% | 0.418 | ||
HY stage 2 | 0.783 | 0.682–0.883 | 68.7% | 60.5% | 79.3% | 73.2% | 68.7% | 87.2% | 0.501 | ||
HY stage 3 | 0.818 | 0.682–0.953 | 76.8% | 50.0% | 76.9% | 76.7% | 60.6% | 91.7% | 0.473 |
