Maternal-fetal transfer rates of PCBs, OCPs, PBDEs, and dioxin-like compounds predicted through quantitative structure-activity relationship modeling.

The present study aims to predict the maternal-fetal transfer rates of the polychlorinated biphenyls (PCBs), organochlorine pesticides (OCPs), and polybrominated diphenyl ethers (PBDEs), and dioxin-like compounds using a quantitative structure-activity relationship model. The relation between the maternal-fetal transfer rate and the contaminants' physicochemical properties was investigated by multiple linear regression (MLR), partial least square regression (PLS), and random forest regression (RF). The 10-fold cross-validation technique estimated low predictive performances for both MLR and PLS models (R 2CV = 0.425 ± 0.0964 for MLR and R 2CV = 0.492 ± 0.115 for PLS) and is in agreement with an external test (R 2pred = 0.129 for MLR and R 2pred = 0.123 for PLS). In contrast, the RF model exhibits good predictive performance, estimated through 10-fold cross-validation (R 2CV = 0.566 ± 0.0885) and an external test set (R 2pred = 0.519). Molecular weight and polarity were selected in all models as important parameters that may predict the ability of a molecule to cross the placenta to the fetus.