All data in this database is collected from the ChemIDplus public database of the TOXNET Databases. They are based on intraperitoneal measurements taken from diverse mouse studies. The whole data set includes 258 diverse organic compounds, which was divided into a training set of 193 compounds and a test set of 65 compounds for regression and classification modeling, respectively. The molecular descriptors of all compounds included in the database files were calculated in the MOE 2009 by fully minimized using the MMFF94x force field before.

Download the databases:

regression_training_set.csv (the SDF flile includes 193 molecules)
regression_test_set.csv (the SDF flile includes 65 molecules)
regression.zip
classification_training_set.csv (the SDF flile includes 193 molecules)
classification_test_set.csv (the SDF file includes 65 molecules)
classification.zip

Reference
[1]. Tailong Lei, Huiyong Sun, Yu Kang, Feng Zhu, Hui Liu, Wenfang Zhou, Zhe Wang, Dan Li, Youyong Li, Tingjun Hou, ADMET Evaluation in Drug Discovery. 18. Reliable Prediction of Chemical-Induced Urinary Tract Toxicity by Boosting Machine Learning Approaches. Molecular Pharmaceutics, 2017. [html] [PDF]