Table 2. Description of the PDBBind datasets.

Version	Refined set	Training set	Core set (test set)	Protein families
v2007	1300	1105	195	65
v2013	2959	2764	195	65
v2015	3706	3511	195	65
v2016	4057	3767	290	58

Number of complexes or number of protein families in PDBBind data sets used in the present binding affinity prediction. Here training sets are set to the corresponding refined sets, excluding the complexes in the corresponding test sets (i.e., core sets). Protein families refer to those in the corresponding core sets.