A systematic review of deep learning chemical language models in recent era

Table 1 Description of databases used in retrieved articles for analysis.

Database	Description	Number of molecules (millions)	Molecule representation	Articles	Ref
PubChem	Structural information of mostly small molecules	115.3	SMILES and InChI	4	[93]
ChEMBL	Bioactive molecules with drug-like properties and Bioactivity records of data	2.4	SMILES and InChI	27	[94]
Zinc	Structural information of drug-like molecules	750	SMILES	27	[95]
US patent database	Reactions extracted by text-mining from United States patents published between 1976 and September 2016	< 1.8	SMILES	1	[96]
DNA-Encoded Library^a	Structural molecular, combinatorial screening, and DNA-encoded information	1040	SMILES	1	[97]
COCONUT	Natural products structural and biological information	0.695	SMILES and InChI	1	[98]
LINCS1000	A comprehensive resource of gene expression in human cells perturbated by small molecules	> 1	Not applicable	1	[99]

^aIndicates databases created by authors and not publicly available, for this case reference indicates the article reference. Number of reported molecules up to September 2024

ISSN: 1758-2946