Table 2.

Key features of the training dataset.

Number of compounds	Active	Inactive	Diversity*	Unique heterocycles			Clusters**	Av. cluster size	Singletons
				All	Active	Inactive
74,567	8,724	65,843	0.86	3,961	1,146	3,370	2,021	15	22,521

*Reverse Tanimoto metrics; **min. cluster size, 5; max. cluster size, 30; similarity threshold, 0.5.