Table 1.

Performance improvement with combined prediction approach in terms of area under the ROC curve (AUC).

Alpha	Average AUC—training ligand data	Average AUC—evaluation ligand data
0	0.591	0.630
0.1	0.635	0.665
0.2	0.675	0.693
0.3	0.710	0.712
0.4	0.738	0.723
0.5	0.759	0.728
0.6	0.774	0.726
0.7	0.779	0.722
0.8	0.778	0.716
0.9	0.774	0.708
1	0.768	0.700

Comparison of the prediction performance when only one scoring method is used (binding-based when α = 0 and cleavage motif-based when α = 1.0) with the combined scoring approach where both binding- and cleavage motif-based scoring schemes were combined together. The AUC was highest for the training data at alpha = 0.7 and highest for evaluation data at alpha = 0.5.