Skip to main content

Table 1 Accuracy metrics for agonist, antagonist, and binding activity cross-validation

From: Exploring non-linear distance metrics in the structure–activity space: QSAR models for human estrogen receptor

Activity # chemicals Model and parameters Sensitivity Specificity Bal accuracy Accuracy ROC AUC Score
Agonist 1538 Morgan kNN arithm k = 10 0.63 0.98 0.80 0.96 0.91 0.70
Agonist 1538 Morgan kNN geom k = 2 0.40 0.99 0.70 0.96 0.73 0.49
Agonist 1538 Morgan kNN exp k = 10 X = 1.5 0.69 0.97 0.83 0.96 0.92 0.73
Agonist 1538 Morgan GkNN k = 10 X = 1 Y = 1 0.63 0.98 0.80 0.96 0.92 0.70
Agonist 1538 Morgan GkNN k = 10 X = 1 Y = 3 0.66 0.97 0.82 0.96 0.92 0.72
Agonist 1538 Morgan GkNN k = 10 X = 1.5 Y = 3 0.74 0.95 0.84 0.94 0.92 0.72
Agonist 1538 Morgan GkNN k = 20 X = 1.5 Y = 5 0.75 0.95 0.85 0.94 0.91 0.73
Antagonist 1645 Morgan kNN arithm k = 3 0.44 1.00 0.72 1.00 0.70 0.51
Antagonist 1645 Morgan kNN geom k = 3 0.00 1.00 0.50 0.99 0.50 0.25
Antagonist 1645 Morgan kNN exp k = 3 X = 1.5 0.44 1.00 0.72 1.00 0.70 0.51
Antagonist 1645 Indigo kNN arithm k = 10 0.22 1.00 0.61 0.99 0.73 0.44
Antagonist 1645 Indigo kNN geom k = 10 0.00 1.00 0.50 0.99 0.50 0.25
Antagonist 1645 Indigo kNN exp k = 10 X = 1.5 0.44 1.00 0.72 0.99 0.73 0.53
Antagonist 1645 Indigo GkNN k = 10 X = 3 Y = 7 0.56 0.98 0.77 0.98 0.73 0.55
Antagonist 1645 Indigo GkNN k = 10 X = 5 Y = 15 0.56 0.98 0.77 0.98 0.73 0.55
Binding 1529 Morgan kNN arithm k = 10 0.63 0.98 0.80 0.96 0.90 0.69
Binding 1529 Morgan kNN geom k = 2 0.43 0.99 0.71 0.96 0.74 0.50
Binding 1529 Morgan kNN exp k = 10 X = 1.5 0.69 0.97 0.83 0.95 0.90 0.71
Binding 1529 Morgan GkNN k = 10 X = 1 Y = 1 0.63 0.98 0.80 0.96 0.90 0.69
Binding 1529 Morgan GkNN k = 10 X = 1 Y = 3 0.66 0.97 0.82 0.95 0.90 0.70
Binding 1529 Morgan GkNN k = 10 X = 1.5 Y = 3 0.73 0.94 0.84 0.93 0.90 0.70
Binding 1529 Morgan GkNN k = 20 X = 1.5 Y = 5 0.75 0.95 0.85 0.94 0.89 0.71
  1. “kNN arithm”, “kNN geom”, and “kNN exp” indicate the kNN models with the arithmetic, geometric, and exponential averaging, respectively. The cumulative score shown in the last column is the product of balanced accuracy, accuracy, and ROC AUC. Italic font indicates accuracy metric values that exceed those for the CERAPP consensus model