TABLE 2.

The overall predictive performance of fingerprint-based models for the test sets.

Model	Test set
	ACC a	F1 b	BA c	SE d	SP e	MCC f	AUC g
DNN::Morgan	0.841 ± 0.059	0.899 ± 0.049	0.703 ± 0.057	0.930 ± 0.034	0.476 ± 0.109	0.468 ± 0.106	0.873 ± 0.043
XGBoost::Morgan	0.854 ± 0.049	0.909 ± 0.042	0.700 ± 0.069	0.947 ± 0.039	0.454 ± 0.155	0.486 ± 0.120	0.861 ± 0.050
SVM::Morgan	0.844 ± 0.063	0.901 ± 0.054	0.692 ± 0.074	0.936 ± 0.051	0.448 ± 0.163	0.453 ± 0.129	0.854 ± 0.050
RF::Morgan	0.849 ± 0.049	0.908 ± 0.041	0.655 ± 0.095	0.965 ± 0.034	0.344 ± 0.212	0.429 ± 0.142	0.870 ± 0.051
LR::Morgan	0.846 ± 0.052	0.901 ± 0.045	0.735 ± 0.047	0.917 ± 0.037	0.554 ± 0.095	0.493 ± 0.085	0.735 ± 0.047
Average (Morgan)	0.847	0.904	0.697	0.939	0.455	0.466	0.839
XGBoost::MACCS	0.850 ± 0.053	0.904 ± 0.049	0.691 ± 0.075	0.939 ± 0.061	0.443 ± 0.189	0.470 ± 0.104	0.853 ± 0.046
DNN::MACCS	0.846 ± 0.038	0.904 ± 0.035	0.681 ± 0.087	0.944 ± 0.032	0.417 ± 0.199	0.448 ± 0.127	0.832 ± 0.041
SVM::MACCS	0.846 ± 0.057	0.903 ± 0.049	0.680 ± 0.067	0.946 ± 0.044	0.414 ± 0.157	0.444 ± 0.107	0.817 ± 0.047
RF::MACCS	0.841 ± 0.062	0.900 ± 0.053	0.673 ± 0.063	0.943 ± 0.049	0.403 ± 0.148	0.442 ± 0.099	0.841 ± 0.044
LR::MACCS	0.835 ± 0.031	0.897 ± 0.028	0.695 ± 0.058	0.920 ± 0.023	0.470 ± 0.117	0.435 ± 0.116	0.695 ± 0.058
Average (MACCS)	0.844	0.902	0.684	0.938	0.429	0.448	0.808

^aACC, accuracy.

^bF1, F1-measure.

^cBA, balanced accuracy.

^dSE, sensitivity.

^eSP, specificity.

^fMCC, matthews correlation coefficient.

^gAUC, the area under receiver operating characteristic.

Bold font illustrates the models that outperformed all other models.