Prediction of compound-target interaction using several artificial intelligence algorithms and comparison with a consensus-based strategy

Table 1 Datasets creation from ChEMBL database

ChEMBL release	Dataset	Process	Compounds	Targets	Interactions
27	CH27	Cleaning	351778	1448	504747
28	CH28	Cleaning	377936	1562	542790
31	CH31	Cleaning	403364	1668	579009
27 & 28	DS1	Training	184046	253\(^{1}\)	249269
31	VSD2	External validation	30526	253\(^{2}\)	42382\(^{3}\)
31	VSD3	Contrast groups	3264	126	4716\(^{3}\)

\(^{1}\) Target has at least 10 chemical interactions, both active and inactive. \(^{2}\) Target has at least 5 chemical interactions, both active and inactive. \(^{3}\) These compound-interactions are only in ChEMBL release 31 but not in 27 or 28

ISSN: 1758-2946