Comparative Studies On Quantitative Predicting Models Of Environmental Toxicity And Skin Permeability For Organic Chemicals

Environmental toxic and skin permeation are important elements for organic chemical import and export inspection and supervison. Here we select several representative groups of compounds with high commercialization, such as aromatic compounds, alcohol, aniline, etc. to carry out research on the non-animal test prediction for environmental acute toxicity and skin permeability.Based on 2D-QSAR and 3D-QSAR, such as group contribution method (GCM), comparative molecular field analysis (CoMFA), comparative molecular similarity indices analysis (CoMSIA) and hologram quantitative structure-activity relationship (HQSAR), a series of predicting models are built up for organic chemicals by different methametic techniques. The detailed results are as follows:1.Two-dimentional strucuture-activity relationships (2D-QSAR) are studied using improved GCM. On one hand, MLR, PLS and ANN models have been developed for predicting the aquatic toxicity (IGC50) of organic chemicals to Tetrahymena pyriformis using UNIFAC groups and 6 physico-chemical properties as decriptors. The introduction of physico-chemical properties improved the correlation coefficient from 0.942 to 0.965, and MSE decreased from 0.113 to 0.071. On the other hand, the addition of strucutural correction factor and octanol-water partition coefficients (logKow) can greatly increased the predicting capability. With these parameters, the QSAR model for 316 aromatic compounds was built up. For training set, the corrlation coefficient r² was0.946, absolute error (AE) and relative error (RE) are 0.346 and 37.3% respectively. While for test set, the corrlation coefficient r² =0.886, AE=0.217, and RE=21.8%2.Three-dimensional quantitative structure-activity relationship (3D-QSAR) studies are carried out based on nineteen N-substituted fluoroacetamides to correlate the acute toxicity in vivo toward rats using comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA). The best CoMFA model included electrostatic and steric fields with the cross-validate regress coefficient Q² of 0.563 and noncross-validate regress coefficient R² of 0.992. CoMSIA analysis includes only steric field is better than CoMFA model with the best result of Q²=0.767.3.The joint toxicity of anilines are tested according to OECD animal test method at different mixture concentrations. With those test results, joint toxicity evaluation is carried out on toxic unit (TU), additional index (AI) and mixture toxic index (MTI) methods to the joint toxic effects. The relationship equations of MTI and mixture concentration are finally built up, which gain better correlation for o-Toluidine and p-Toluidine, aniline and 2,4-Dimethyl aniline (R²>0.9).4.Using theoretical molecular descriptors, two optimal QSPR models were built up for skin permeability of 82 compounds such as saturated alcohol, actone, acid, etc. Data set was divided into training set and test set by K-means clustering technique to gain statistically robust QSPR models. Compared to the stepwise MLR, the model developed by PLS had higher correlative but lower predictive capabilities in terms of Q² (0.878) and R²_pred (0.892). The quantitative structure-activity relationship model for the Franze cell permeabilities of 44 compounds was obtained by using HQSAR methotology, in which types of chemical bond and Hydrobond donor-acceptor are the two main influencing factor. The model got the Q² of 0.682 and SEEcv of 0.536, noncross-validate regress coefficient R² of 0.950 and SEE of 0.213.