SMPLIP-Score: predicting ligand binding affinity from simple and interpretable on-the-fly interaction fingerprint pattern descriptors

Table 5 The prediction performance on benchmark datasets and statistical comparison of SMPLIP-Score with reported models

Datasets	Models	SETS	PCC	RMSE	MAE	p_Value	Sp	Refs
Astex Diverse Set	SMPLIP-Score	–	0.724	1.177	0.938	0.002	0.764	–
CSAR NRC HiQ		Set01	0.785	1.903	1.500	8.03E−13	0.761
CSAR NRC HiQ		Set02	0.803	1.475	1.134	1.54E−15	0.823
FEP		BACE	0.239	0.639	0.505	0.160	0.250
		MCL1	0.077	1.045	0.797	0.629	0.146
		PTP1B	0.634	0.768	0.536	0.002	0.680
		Thrombin	− 0.645	0.962	0.780	0.321	− 0.536
		Tyk2	0.469	0.859	0.655	0.078	0.546
PDBbind NMR		–	0.209	1.857	1.552	0.004	0.234
Astex Diverse Set	DeepAtom	–	0.768	1.027	0.714	–	–	Li et al. [25]
	RF-Score	–	0.710	1.144	0.891	–	–	Ballester et al. [63]
	Pafnucy	–	0.569	1.374	1.110	–	–	Stepniewska Dziubinska et al. [28]
	Res4HTMD	–	–	1.54	–	0.07	0.41	Hassan Harrirou et al. [29]
	RosENet	–	–	1.84	–	0.21	0.29	Hassan Harrirou et al. [29]
CSAR NRC HiQ	K_DEPP	Set01	0.72	2.09	–	–	–	Jiménez et al. [24]
	K_DEPP	Set02	0.65	1.92	–	–	–	Jiménez et al. [24]
	Res4HTMD	Set01	–	1.75	–	2E−15	0.84	Hassan Harrirou et al. [29]
	Res4HTMD	Set02	–	1.34	–	3E−13	0.83
	RosENet	Set01	–	1.71	–	2E−17	0.87
	RosENet	Set02	–	1.38	–	2E−14	0.85
FEP	K_DEEP	BACE	− 0.06	0.84	–	–	–	Jiménez et al. [24]
		MCL1	0.34	1.04	–	–	–
		PTP1B	0.58	0.93	–	–	–
		Thrombin	0.58	0.44	–	–	–
		Tyk2	− 0.22	1.13	–	–	–
	Res4HTMD	BACE	–	1.27	–	0.26	− 0.19	Hassan Harrirou et al. [29]
		MCL1	–	1.1	–	2E−3	0.45
		PTP1B	–	0.88	–	6E−3	0.55
		Thrombin	–	0.83	–	0.16	0.45
		Tyk2	–	0.76	–	2E−3	0.71
PDBbind NMR	RosENet	–	–	1.37	–	–	0.56

Models ML/DL method used to build the ligand binding affinity prediction model, RMSE root-mean-square-error, MAE mean absolute error, PCC Pearson correlation coefficient, p_value p_value for statistical significance, Sp Spearman correlation coefficient, RF-Model RF parameters includes: n_estimators = 500; max_features = “AUTO”

ISSN: 1758-2946