TeratoDB

These files contain a set of 585 molecules classified for teratogenicity, with structural and mechanism of action data. The dataset has been balanced as much as possible with the data available in the scientific literature and public repositories like the EPA's IRIS and the NLM's TOXNET Hazardous Substance Data Bank. The goal of this dataset is to collect the available data in a format that will allow for easy prototyping of machine learning applications for teratogenic toxicity.

The file 'drug_data.xlsx' contains the list of molecules with an assigned teratogenicity category (1 for teratogenic agents, -1 for non-teratogenic agents). It also contains ChEMBL, DrugBank and PubChem identifiers, when available, and a SMILES representation of the molecule. This list has been curated by hand.
The file 'drug_uniprot_targets.xlsx', contains 2290 drug-target relationships between the 585 drugs from the previous file and 757 molecular targets that were found to be annotated in ChEMBL and DrugBank. The files used to process said data have been included in 'target_processing.zip'. Uniprot identifiers were used for all of the targets. Drug-target relationships found in both databases were included only once. Information about the Mechanism of Action was included when available.

Finally, the file called 'target_list.xlsx' contains the list of the unique 758 protein targets found in any or both of the databases. They include Uniprot and ChEMBL identifiers, and a description of said targets when available.