CytoToxLCM: A Software to Predict Cytotoxicity of Emerging Contaminant Liquid Crystal Monomers
Jia Qian Wu, Chengxia Bian, Xianhai Yang, Guanyong Su
Abstract
Liquid crystal monomers (LCMs) have emerged as a new class of contaminants, yet their health risks remain unclear due to limited toxicity data. This study assessed 46 structurally diverse LCMs using primary mouse hepatocytes, revealing significant cytotoxicity in 22 compounds, particularly 3OCB, tFMeO-3cHtFT, 2OdF3B, 2O2cHdFB, and 2CB. To predict cytotoxicity across thousands of reported LCMs, classification models and quantitative structure–activity relationship (QSAR) models were developed. For five optimal classification models, their sensitivity, specificity, and prediction accuracy for the respective training and validation sets were >0.900. In terms of quantitative prediction, we established a k-nearest neighbor-based QSAR model, and its coefficient of determination ( R 2 ), leave-one-out cross-validation Q 2 ( Q 2 LOO ), externally explained variance ( Q 2 EXT ), and concordance correlation coefficient ( CCC ) were all greater than 0.850. These models were integrated into CytoToxLCM software, enabling the high-throughput screening of 1127 LCMs. The results of virtual screening showed that over 40% of the 1127 LCMs were predicted to be cytotoxic, with fluorinated LCMs ranking as the most toxic. The models established in this study are reliable for predicting the cytotoxicity of new or untested LCMs, aiding in better understanding their potential hepatotoxic effects and contributing to the design of safer industrial application alternatives.