Datasets and codes for the recreation and deployment of predictive toxicology models developed to predict 79 human molecular initiating events (MIEs) using neural networks (DOI: 10.1039/D0SC01637C).

Python codes are included to generate molecular fingerprints, generate chemical clusters, build models and recall those models.

Model files generated during clustered cross-validation and for comparison to structural alerts and random forests are also included.

Input biological data used in model construction is included for reference as complete datasets (Total), files divided as training and test for comparison to other models (Comparison), and clustered files used in clustered cross-validation (Clustered).