TABLE 2.
Performance metrics of adenoviral infection prediction models. RBF, Radial Basis Function; AUC, Area Under the Curve; MCC, Matthew’s Correlation Coefficient.
| Host taxa level | Classifier | Sensitivity | Specificity | Accuracy | F-score | MCC | AUC |
|---|---|---|---|---|---|---|---|
| Genus | SVM (kernel = “rbf,” gamma = “auto”) | 0.92 ± 0.009 | 0.86 ± 0.047 | 0.92 ± 0.009 | 0.96 ± 0.005 | 0.39 ± 0.035 | 0.89 ± 0.023 |
| MLP (activation = “tanh,” hidden layer=(16,4)) | 0.95 ± 0.009 | 0.73 ± 0.064 | 0.94 ± 0.009 | 0.97 ± 0.005 | 0.40 ± 0.045 | 0.84 ± 0.031 | |
| Random forest (number of trees = 50, criterion = “entropy,” max_depth = 16) | 0.96 ± 0.006 | 0.70 ± 0.071 | 0.95 ± 0.006 | 0.98 ± 0.003 | 0.42 ± 0.043 | 0.83 ± 0.035 | |
| Family | SVM (kernel = “rbf,” gamma = “auto”) | 0.91 ± 0.009 | 0.84 ± 0.057 | 0.91 ± 0.008 | 0.95 ± 0.005 | 0.36 ± 0.031 | 0.88 ± 0.027 |
| MLP (activation = “tanh,” hidden layer=(16,4)) | 0.94 ± 0.012 | 0.72 ± 0.064 | 0.94 ± 0.011 | 0.97 ± 0.006 | 0.37 ± 0.048 | 0.83 ± 0.031 | |
| Random forest (number of trees = 50, criterion = “entropy”, max_depth = 16) | 0.95 ± 0.007 | 0.66 ± 0.068 | 0.95 ± 0.007 | 0.97 ± 0.004 | 0.38 ± 0.043 | 0.81 ± 0.033 | |
| Order | SVM (kernel = “rbf,” gamma = “auto”) | 0.91 ± 0.009 | 0.82 ± 0.057 | 0.90 ± 0.009 | 0.95 ± 0.005 | 0.34 ± 0.028 | 0.86 ± 0.027 |
| MLP (activation = “tanh,” hidden layer=(16,4)) | 0.94 ± 0.011 | 0.70 ± 0.075 | 0.94 ± 0.011 | 0.97 ± 0.006 | 0.37 ± 0.051 | 0.82 ± 0.038 | |
| Random forest (number of trees = 50, criterion = “entropy,” max_depth = 16) | 0.95 ± 0.007 | 0.66 ± 0.070 | 0.95 ± 0.007 | 0.97 ± 0.004 | 0.37 ± 0.040 | 0.81 ± 0.034 | |
| Class | SVM (kernel = “rbf,” gamma = “auto”) | 0.88 ± 0.011 | 0.82 ± 0.061 | 0.88 ± 0.010 | 0.93 ± 0.006 | 0.30 ± 0.028 | 0.85 ± 0.029 |
| MLP (activation = “tanh,” hidden layer=(16,4)) | 0.94 ± 0.010 | 0.68 ± 0.067 | 0.93 ± 0.010 | 0.97 ± 0.005 | 0.35 ± 0.043 | 0.81 ± 0.032 | |
| Random forest (number of trees = 50, criterion = “entropy,” max_depth = 16) | 0.95 ± 0.007 | 0.63 ± 0.071 | 0.94 ± 0.007 | 0.97 ± 0.004 | 0.35 ± 0.043 | 0.79 ± 0.035 | |
| None | SVM (kernel = “rbf,” gamma = “auto”) | 0.88 ± 0.011 | 0.83 ± 0.064 | 0.88 ± 0.010 | 0.93 ± 0.006 | 0.30 ± 0.029 | 0.86 ± 0.030 |
| MLP (activation = “tanh,” hidden layer=(16,4)) | 0.94 ± 0.009 | 0.68 ± 0.079 | 0.93 ± 0.008 | 0.96 ± 0.005 | 0.34 ± 0.043 | 0.81 ± 0.038 | |
| Random forest (number of trees = 50, criterion = “entropy,” max_depth = 16) | 0.95 ± 0.007 | 0.63 ± 0.072 | 0.94 ± 0.007 | 0.97 ± 0.004 | 0.35 ± 0.041 | 0.79 ± 0.035 |
Bolded value indicates the implementation of the SVM algorithm yielded the best performance in terms of sensitivity for infection prediction for our particular dataset for all the experiments.