Table 3 Performance metrics for the CPDB dataset
From: An ensemble model of QSAR tools for regulatory risk assessment
Model | Accuracy (%) | SN (%) | SP (%) | BA (%) | PPV (%) | NPV (%) | Kappa (\(\kappa\)) |
|---|---|---|---|---|---|---|---|
Toxtree | 66.04 | 84.50 | 44.59 | 64.55 | 63.93 | 71.22 | 0.30 |
Lazar | 80.63 | 86.05 | 74.32 | 80.19 | 79.57 | 82.09 | 0.61 |
Danish QSAR | 65.00 | 91.09 | 34.68 | 62.89 | 61.84 | 77.00 | 0.27 |
OECD toolbox | 64.79 | 84.50 | 41.89 | 63.20 | 62.82 | 69.93 | 0.27 |
Bayes ensemble (Cut-off = 0.4) | 81.04 | 83.33 | 75.23 | 79.28 | 80.14 | 82.27 | 0.62 |
Bayes ensemble (Cut-off = 0.5) | 80.21 | 84.50 | 75.23 | 79.87 | 79.85 | 80.68 | 0.60 |
Bayes ensemble (Cut-off = 0.6) | 80.42 | 84.50 | 77.03 | 80.77 | 80.83 | 79.91 | 0.61 |