Table 4.

Performances with LRs using DeepSnap_python system of the three MIE targets.

PubChem		720725_GR_Ant	1347030_TRHR	1347032_TGF_Beta
Assay AID		Train:Valid:Test = 7:1:2	Train:Valid:Test = 3:1:2	Train:Valid:Test = 7:1:2
ROC_AUC	average	0.884 ± 0.053	0.897 ± 0.016	0.909 ± 0.011
	max_ROC_AUC	0.930	0.911	0.922
	max_LR	0.00009	0.000002	0.000021
BAC	average	0.817 ± 0.053	0.844 ± 0.012	0.839 ± 0.010
	max_BAC	0.865	0.865	0.853
	max_LR	0.0007	0.000001	0.000029
MCC	average	0.354 ± 0.090	0.171 ± 0.015	0.361 ± 0.016
	max_MCC	0.466	0.191	0.387
	max_LR	0.00007	0.0048	0.000029
Acc	average	0.859 ± 0.060	0.811 ± 0.025	0.807 ± 0.028
	max_Acc	0.928	0.848	0.855
	max_LR	0.00007	0.000005	0.00002
loss_train	average	0.215 ± 0.231	0.098 ± 0.062	0.125 ± 0.110
	min_loss	0.022	0.020	0.038
	min_LR	0.00003	0.00002	0.00003
loss_val	average	0.263 ± 0.186	0.122 ± 0.058	0.236 ± 0.062
	min_loss	0.124	0.066	0.170
	min_LR	0.00003	0.0008	0.000021
PR_AUC	average	0.502 ± 0.177	0.155 ± 0.045	0.410 ± 0.064
	max_PR_AUC	0.789	0.213	0.472
	max_LR	0.00007	0.0042	0.00003
F	average	0.898 ± 0.039	0.886 ± 0.015	0.858 ± 0.019
	max_F	0.942	0.909	0.890
	max_LR	0.00007	0.000005	0.00002