The data sets accompany the manuscript 10.1101/2023.11.15.567095 The data sets rely on previous work from M. Lovrić, "Razvoj i primjena modela za procjenu ekotoksikoloških rizika bioaktivnih kemijskih spojeva", Disertacija, Sveučilište u Zagrebu, Prirodoslovno-matematički fakultet, Zagreb, 2021. Dostupno na: https://urn.nsk.hr/urn:nbn:hr:217:730117 and Lovrić M, Malev O, Klobučar G, Kern R, Liu JJ, Lučić B. Predictive Capability of QSAR Models Based on the CompTox Zebrafish Embryo Assays: An Imbalanced Classification Problem. Molecules. 2021 Mar 15;26(6):1617. doi: 10.3390/molecules26061617 the original data was downloaded from the Tox21 challenge https://tripod.nih.gov/tox21/challenge/data.jsp The files comprise of: tox21_desc_p194.csv - descriptor data for the prepared Tox21 data set tox21_fp_all.csv - fingerprint and maccs data for the prepared Tox21 data set tox21_target_all.csv - 12 binary Tox21 target(labels) standardizer_pipeline.xml - structure standardizer pipeline for ChemAxon Standardizer tox21_models_metadata.json - model metadata such as sklearn hyperparameters and descriptor feature lists