Table 5. Performance of prediction and chemical similarities ^a,b.

Model	Training 1 vs Test 1	Training 2 vs Test 2	Training 3 vs 75 A2A ligands	Training 3 vs 56 drugs b
# of clusters in training	26	55	55	55
# of actives in training	53	35	97	97
# of inactives in training	88	48	171	171
Max similarity to test set c	0.65c	0.67	0.37	0.13
Sensitivity, %	82	96	78	72
Specificity, %	94	94	74	77

^a Sensitivity = TP/(TP+FN), specificity = TN/(TN+FP), false positive rate = FP/(FP + TN), false negative rate = FN/(TP+FN), where TP is true positive, TN is true negative, FP is false positive, FN is false negative.

^b Based on in vitro assay results, TP = 5, TN = 38, FP = 11, FN = 2.

^c Similarities calculated using radial binary fingerprints. The 268 training and test compounds were represented by 55 centroid structures from the 55 chemical clusters.