Exploration and augmentation of pharmacological space via adversarial auto-encoder model for facilitating kinase-centric drug development

Table 2 Summary of ENB and EPA performance on external datasets

Dataset	Method	Regression metrics			Classification metrics
Dataset	Method	PCC	MAE	MSE	AUC	F1 score	Precision	Recall
Metz	ENB	0.47	0.51	0.49	0.81	0.45	0.39	0.54
Metz	EPA	0.54	0.44	0.39	0.84	0.49	0.41	0.61
Davis	ENB	0.38	1.37	0.90	0.67	0.58	0.77	0.46
Davis	EPA	0.42	1.10	0.79	0.69	0.64	0.76	0.55
PKIS1	ENB	0.29	0.64	0.59	0.75	0.25	0.21	0.30
PKIS1	EPA	0.33	0.56	0.44	0.79	0.26	0.22	0.32
KinaseSafari	ENB	0.33	1.1	2.14	0.68	0.56	0.74	0.45
KinaseSafari	EPA	0.44	1.01	1.72	0.73	0.61	0.77	0.51
MRC	ENB	0.38	0.83	1.14	0.68	0.33	0.53	0.24
MRC	EPA	0.45	0.78	0.99	0.74	0.38	0.54	0.30

The bold number denotes the better result between ENB and EPA for predicting the corresponding external dataset

ISSN: 1758-2946