TABLE 1.
Representative biochemical datasets used in deep learning studies.
| Dataset | Description | URL | References |
| ZINC | ZINC database contains over 230 million compounds. | http://zinc.docking.org/ | Bai et al., 2020; Choi et al., 2020; Stokes et al., 2020; Ton et al., 2020 |
| ChEMBL | ChEMBL (version 27) chemical database contains over 1.9 million specific compounds. | https://www.ebi.ac.uk/chembl/ | Stokes et al., 2020 |
| Drug target commons (DTC) | DTC crowdsourcing database contains 204,901 annotated bioactivity data points among 4,276 compounds and 1,007 specific protein targets. | https://drugtargetcommons.fimm.fi/ | Beck et al., 2020 |
| BindingDB | BindingDB database of measured binding affinities contains 2,061,017 binding data for 8,160 protein targets and 907,259 small molecules. | http://www.bindingdb.org/bind/index.jsp | Beck et al., 2020 |
| DrugBank | DrugBank pharmaceutical database contains detailed molecular information about drugs, their mechanisms, interactions and targets. | https://go.drugbank.com/releases/latest | Choi et al., 2020; Zeng et al., 2020 |
| PDBbind | PDBbind database provides binding data of 21,382 biomolecular complexes, including protein-ligand (17,679), nucleic acid-ligand (136), protein-nucleic acid (973), and protein-protein complexes (2,594). | http://www.pdbbind.org.cn | Bai et al., 2020 |