Distance plus attention for binding affinity prediction

Table 3 Results of five models and their averages when each model is trained on distinct subsets of the training dataset (PDBbind2016 and PDBbind2020) for the CASF\(-\)2016.290

Training Dataset	Predictors	R\((\uparrow )\)	RMSE\((\downarrow )\)	MAE\((\downarrow )\)	SD\((\downarrow )\)	CI\((\uparrow )\)
PDBbind2016	Model-1	0.872	1.094	0.855	1.096	0.843
	Model-2	0.857	1.168	0.931	1.170	0.831
	Model-3	0.853	1.153	0.906	1.155	0.828
	Model-4	0.862	1.138	0.895	1.140	0.838
	Model-5	0.853	1.183	0.935	1.185	0.829
	Ensemble	0.888	1.061	0.827	1.063	0.853
PDBbind2020	Model-1	0.878	1.106	0.858	1.108	0.852
	Model-2	0.869	1.128	0.838	1.130	0.847
	Model-3	0.859	1.134	0.840	1.136	0.840
	Model-4	0.852	1.193	0.935	1.195	0.826
	Model-5	0.856	1.227	0.939	1.230	0.828
	Ensemble	0.908	0.999	0.732	1.001	0.874

Higher R and CI values and lower RMSE, MAE, and SD values denote superior performances. The best-performing values are emboldened

ISSN: 1758-2946