Results of external validation of Bayesian model for DILI
The results were for the complete test set, true positive (TP) = 86, true negative (TN) =56, false positive (FP) = 28, and false negative (FN) = 67. For the subset of most interest, TP = 13, TN = 10, FP = 5, and FN = 8. For the compounds >70% similar to the training set, TP = 9, TN = 8, FP = 2, and FN = 9. Matthews correlation coefficient = (TP × TN − FP × FN)/((TP + FN)(TP + FP)(TN + FP)(TN + FN)) · 0.5. Concordance (prediction accuracy) = (TP + TN)/(TP + TN + FP + FN), specificity = TN/(TN + FP), and sensitivity = TP/(TP + FN).
| Test Set | Matthews Correlation Coefficient | Concordance | Specificity | Sensitivity |
|---|---|---|---|---|
| % | % | % | ||
| Complete test set (n = 237) | 0.22 | 59.91 | 66.67 | 56 |
| Subset of most interest (n = 37) | 0.28 | 63.88 | 66.67 | 61.9 |
| Compounds >70% similar to training set (n = 28) | 0.29 | 60.71 | 80.00 | 50 |