(567d) Integrated Prediction of Chemical Effects on Important Pathways of Embryo Development Based on in Vitro Embryonic Stem Cell Test and Machine Learning

Currently, there are thousands of chemicals in common use; however, only a small number of them have sufficient toxicity evaluation, and even less information for the specific embryo-related toxicity field: embryotoxicity and developmental toxicity. Embryonic stem cell test (EST) is the only accepted in vitromethod for assessing embryotoxicity without animal sacrifice. However, EST for regulatory embryotoxicity screening are impeded by its complexity, time-consuming, limited differentiation lineage endpoints, and poor availability for use to develop a reliable prediction model (PM) for embryotoxicity assessments.

A key step in the understanding of embryo development is to construct a systematic and comprehensive model of network with the development-related pathways involved. In our research, we mainly focus on the essential markers in the pathways related to ESCs apoptosis and proliferation, pluripotency, and specific-lineage differentiation. Quantitative structure-activity relationship (QSAR) models which predict the biological activities from chemical structures information is an efficient and less time-consuming strategy compared to experimental-based studies. Therefore, an integrated prediction pipeline for in vitrohigh throughput screening (HTS) based on engineered murine embryonic stem cells (mESCs) expressing a reporter, enhanced green fluorescent protein (EGFP), driven by human promoters for key genes in the pathways associated with apoptosis, pluripotency, and tissue-specific differentiation cultured in novel multi-well plates is developed and used to obtain new embryotoxicity and developmental toxicity data needed for the training of machine learning models such as support vector machines (SVM) and deep neural network (DNN). The application of advanced machine learning algorithms can significantly improve the accuracy and efficiency of the PM for embryotoxicity assessment of compounds with unknown developmental toxicity. This study will provide the technology needed for fast and accurate embryotoxicity assessments. The data obtained from our integrated analysis can provide the information needed for the regulation of industrial and emerging chemicals.