Optimization of the TeraTox assay for preclinical teratogenicity assessment

Abstract Current animal-free methods to assess teratogenicity of drugs under development still deliver high numbers of false negatives, and more sensitive approaches of toxicity prediction are required. To address this issue, we characterized the TeraTox test, a newly developed multi-lineage differentiation assay for human teratogenicity prediction using 3D human induced pluripotent stem cells. TeraTox produces as primary output concentration-dependent data sets for each test compound on cytotoxicity and altered gene expression. These data are then fed into a prediction model based on an interpretable machine-learning approach. The final information obtained relates to the concentration-dependent human teratogenicity potential of drug candidates. We applied TeraTox to profile 33 approved pharmaceuticals and 12 proprietary drug candidates with known in vivo data. This way, it was possible to relate the test predictions to known human or animal toxicity. The TeraTox had an accuracy of 69% (specificity: 53%, sensitivity: 79%). It clearly performed better than two quantitative structure-activity relationship (QSAR) models and it had a higher sensitivity than the murine embryonic stem cell test (mEST) run in the same laboratory. By combining TeraTox and mEST data, the overall prediction accuracy was further improved. The knowledge on the pattern of altered gene expression may provide additional value in grouping toxicologically similar compounds and possibly deducing common modes of action. The assay will thus be a helpful additional tool in drug discovery, and the dataset provided here will be a valuable resource for the field of stem cell-based drug profiling.