Research data supporting: "Machine learning in predictive toxicology: investigating developmental and reproductive toxicity with transfer learning". The dataset contains a total of 3245 chemicals related to DART (1662 positives, 1583 negatives) and were taken from several sources in the literature, including both in vitro and in vivo data. All SMILES for the compounds have salts and metals removed. For more details on the data sources see Table 10 in Chapter 4 of the thesis or the README sheet in the .xlsx file.