DeepAR: a novel deep learning-based hybrid framework for the interpretable prediction of androgen receptor antagonists

Table 1 Summary of twelve molecular fingerprints used in this study

Fingerprint	Abbreviation	#Feature	Description	Ref.
2D atom pair	AP2D	780	Presence of atom pairs at various topological distances	[83]
CDK	CKD	1024	Fingerprint of length 1,024 and search depth of 8	[84]
CDK extended	CKDExt	1024	Extends the fingerprint with additional bits describing ring features	[84]
CDK graph only	CKDGraph	1024	A special version that considers only the connectivity and not bond order	[84]
Circle	Circle	1024	Circular fingerprint	[85]
EState	EState	79	Electrotopological state atom types	[86]
Hybrid	Hybrid	1024	CDK hybridization fingerprint	[85]
Klekota–Roth	KR	4860	Presence of chemical substructures	[87]
MACCS	MACCS	166	Binary representation of chemical features defined by MACCS keys	[88]
Pubchem	Pubchem	881	Binary representation of substructures defined by PubChem	[89]
Substructure	FP4	307	Presence of SMARTS patterns for functional groups	[90]
Substructure count	FP4C	307	Count of SMARTS patterns for functional groups	[90]

ISSN: 1758-2946