Estimation of the applicability domain of kernel-based machine learning models for virtual screening

Table 7 Ranges of the applicability score for the different combinations of ADE formulation, kernel and experiment

ADE	Kernel	Target	Training Set			Screening Set
			Min	Max	Avg.	Min	Max	Avg.
KDE	OAK	Thrombin	0.70	0.87	0.82 ± 0.04	0.51	0.77	0.64 ± 0.08
		Factor Xa	0.59	0.75	0.71 ± 0.02	0.47	0.72	0.60 ± 0.08
		PDGFRβ	0.81	0.88	0.85 ± 0.02	0.54	0.70	0.62 ± 0.05
	FlexOAK	Thrombin	0.69	0.86	0.81 ± 0.04	0.54	0.79	0.66 ± 0.08
		Factor Xa	0.60	0.74	0.70 ± 0.02	0.53	0.69	0.61 ± 0.05
		PDGFRβ	0.80	0.88	0.85 ± 0.02	0.54	0.68	0.61 ± 0.04
	MARG	Thrombin	0.81	0.95	0.92 ± 0.03	0.57	0.91	0.74 ± 0.11
		Factor Xa	0.74	0.94	0.91 ± 0.02	0.52	0.88	0.70 ± 0.11
		PDGFRβ	0.92	0.96	0.95 ± 0.01	0.74	0.91	0.82 ± 0.05
wKDE	OAK	Thrombin	0.70	0.87	0.82 ± 0.04	0.51	0.79	0.65 ± 0.09
		Factor Xa	0.59	0.76	0.71 ± 0.02	0.47	0.71	0.59 ± 0.07
		PDGFRβ	0.81	0.88	0.85 ± 0.02	0.54	0.70	0.62 ± 0.05
	FlexOAK	Thrombin	0.69	0.87	0.82 ± 0.04	0.54	0.79	0.66 ± 0.08
		Factor Xa	0.60	0.74	0.71 ± 0.02	0.53	0.70	0.62 ± 0.05
		PDGFRβ	0.80	0.89	0.85 ± 0.02	0.54	0.68	0.61 ± 0.04
	MARG	Thrombin	0.80	0.96	0.93 ± 0.04	0.56	0.93	0.75 ± 0.12
		Factor Xa	0.74	0.94	0.91 ± 0.02	0.52	0.87	0.69 ± 0.11
		PDGFRβ	0.91	0.97	0.95 ± 0.01	0.72	0.91	0.82 ± 0.06
One-Class	OAK	Thrombin	-6.52	1.61	-0.33 ± 1.51	-15.70	-0.79	-8.25 ± 4.65
		Factor Xa	-22.21	4.98	-0.90 ± 3.41	-55.36	-3.94	-29.65 ± 16.01
		PDGFRβ	-1.14	0.92	0.0 ± 0.43	-17.08	-6.63	-11.86 ± 3.24
	FlexOAK	Thrombin	-5.62	1.35	-0.37 ± 2.07	-16.87	-2.42	-9.65 ± 4.50
		Factor Xa	-20.2.0	3.52	-0.96 ± 2.96	-38.80	-3.68	-21.24 ± 10.94
		PDGFRβ	-1.41	0.61	-0.12 ± 0.47	-17.01	-8.46	-12.74 ± 2.66
	MARG	Thrombin	-5.22	1.04	-0.35 ± 1.21	-20.82	-0.03	-10.42 ± 6.45
		Factor Xa	-26.29	2.85	-0.77 ± 3.39	-27.55	0.00	-13.78 ± 8.55
		PDGFRβ	-1.11	0.50	-0.01 ± 0.31	-9.56	-1.37	-5.47 ± 2.54

ISSN: 1758-2946