Table 2.
Performance metrics of the best models for each metabolic subphenotype for each feature set
| Prediction task | Feature set | auROC | Sensitivity | Specificity | F1 | Precision | Accuracy | Classifier |
|---|---|---|---|---|---|---|---|---|
| Muscle IR | Demographics | 0.37 | 0.55 | 0.56 | 0.55 | 0.58 | 0.56 | SVM-RBF |
| Demographics + PRS | 0.21 | 0.38 | 0.26 | 0.27 | 0.25 | 0.29 | RF | |
| Demographics + Lab | 0.69 | 0.67 | 0.65 | 0.66 | 0.68 | 0.66 | LR-L1 | |
| Demographics + Incretins | 0.38 | 0.85 | 0.34 | 0.66 | 0.58 | 0.59 | SVM-RBF | |
| Demographics + HOMA-B | 0.44 | 0.98 | 0.02 | 0.40 | 0.50 | 0.53 | LR-L1 | |
| Demographics + HOMA-IR | 0.77 | 0.74 | 0.71 | 0.73 | 0.76 | 0.72 | LR-L1 | |
| Demographics + Matsuda | 0.83 | 0.85 | 0.42 | 0.71 | 0.66 | 0.65 | SVM-linear | |
| Demographics + OGTT_G_Curve_Features | 0.90 | 0.90 | 0.75 | 0.83 | 0.81 | 0.82 | SVM-linear | |
| Demographics + OGTT_G_ReducedRep | 0.95 | 0.81 | 0.91 | 0.85 | 0.93 | 0.87 | LR-L1 | |
| β-cell function | Demographics | 0.49 | 0.69 | 0.30 | 0.52 | 0.60 | 0.51 | SVM-RBF |
| Demographics + PRS | 0.46 | 0.77 | 0.12 | 0.35 | 0.42 | 0.44 | LR-L1 | |
| Demographics + Lab | 0.50 | 0.82 | 0.18 | 0.59 | 0.62 | 0.56 | SVM-linear | |
| Demographics + Incretins | 0.72 | 0.93 | 0.01 | 0.54 | 0.43 | 0.53 | SVM-RBF | |
| Demographics + HOMA-B | 0.48 | 0.98 | 0.01 | 0.52 | 0.22 | 0.58 | SVM-linear | |
| Demographics + HOMA-IR | 0.57 | 0.71 | 0.19 | 0.55 | 0.55 | 0.50 | LR-L1 | |
| Demographics + Matsuda | 0.44 | 0.74 | 0.29 | 0.61 | 0.62 | 0.56 | SVM-RBF | |
| Demographics + OGTT_G_Curve_Features | 0.84 | 0.78 | 0.72 | 0.79 | 0.82 | 0.75 | SVM-linear | |
| Demographics + OGTT_G_ReducedRep | 0.89 | 0.82 | 0.73 | 0.82 | 0.85 | 0.78 | SVM-linear | |
| Incretin effect | Demographics | 0.62 | 0.51 | 0.88 | 0.63 | 0.81 | 0.72 | LR-L1 |
| Demographics + PRS | 0.53 | 0.33 | 0.67 | 0.37 | 0.41 | 0.54 | RF | |
| Demographics + Lab | 0.80 | 0.64 | 0.78 | 0.66 | 0.74 | 0.71 | RF | |
| Demographics + Incretins | 0.68 | 0.60 | 0.64 | 0.54 | 0.56 | 0.62 | RF | |
| Demographics + HOMA-B | 0.68 | 0.62 | 0.61 | 0.55 | 0.57 | 0.61 | RF | |
| Demographics + HOMA-IR | 0.75 | 0.60 | 0.76 | 0.63 | 0.71 | 0.69 | RF | |
| Demographics + Matsuda | 0.48 | 0.72 | 0.56 | 0.63 | 0.56 | 0.62 | SVM-linear | |
| Demographics + OGTT_G_Curve_Features | 0.90 | 0.75 | 0.74 | 0.71 | 0.73 | 0.74 | SVM-RBF | |
| Demographics + OGTT_G_ReducedRep | 0.88 | 0.79 | 0.72 | 0.73 | 0.71 | 0.75 | LR-L1 | |
| Hepatic IR | Demographics | 0.63 | 0.71 | 0.47 | 0.60 | 0.57 | 0.58 | LR-L1 |
| Demographics + PRS | 0.45 | 0.34 | 0.61 | 0.33 | 0.36 | 0.51 | SVM-RBF | |
| Demographics + Lab | 0.78 | 0.90 | 0.45 | 0.74 | 0.66 | 0.67 | SVM-RBF | |
| Demographics + Incretins | 0.73 | 0.70 | 0.60 | 0.63 | 0.62 | 0.64 | RF | |
| Demographics + HOMA-B | 0.50 | 0.52 | 0.48 | 0.48 | 0.50 | 0.48 | SVM-RBF | |
| Demographics + HOMA-IR | 0.88 | 0.80 | 0.80 | 0.78 | 0.83 | 0.80 | LR-L1 | |
| Demographics + Matsuda | 0.90 | 0.82 | 0.71 | 0.76 | 0.78 | 0.76 | SVM-linear | |
| Demographics + OGTT_G_Curve_Features | 0.84 | 0.78 | 0.73 | 0.76 | 0.76 | 0.75 | RF | |
| Demographics + OGTT_G_ReducedRep | 0.54 | 0.73 | 0.52 | 0.65 | 0.64 | 0.63 | SVM-linear |