卤代脂肪烃鱼类急性毒性QSAR模型研究

QSAR STUDY ON THE ACUTE TOXICITY OF ALKYL HALIDES TO PIMEPHALES PROMELAS

采用量子化学方法对27个卤代烃化合物进行结构优化和频率分析计算,提取其组成、拓扑、几何、静电和量子化学等结构参数描述符,运用启发式方法筛选最佳的结构参数,并建立黑头呆鱼急性毒性线性回归模型和非线性的支持向量机回归模型.模型方程表明卤代烃化合物导致黑头呆鱼的急性毒性可以应用KH1、IC0、EHOMO-LUMO三个参数解释,几何形状,疏水性能和反应活性是影响鱼类急性毒性的主要因素.两个模型均具有较强的模型稳定性、预测性和可信性,可用于准确预测卤代烃在鱼体内的急性毒性.

The DFT-B3LYP method,with the basis sets 6-311＋g（d,p）,was employed to calculate the molecular geometries and electronic structures,and to analyze frequency of 27 alkyl halides.Then 282 molecular descriptors were calculated,divided into compositional,topological,geometrical,electrostatic,quantum-chemical and thermodynamic parameters.Heuristic strategies was used for the effective search of the best descriptors in the large space of the natural descriptors,then established the quantitative structure-activity relationship （QSAR） by multivariate linear regressions（MLR） and support vector machine（SVM）.Three molecular descriptors（KH1,IC0,EHOMO-LUMO） were found to relate significantly to acute toxicities.The proposed QSAR models by MLR and SVM are robust and satisfactory,the values predicted by the two model are in good agreement with the experiment values.All the descriptors utilized are calculated solely from the structures of the molecules,which makes it feasible to predict LC50 of alkyl halides to Pimephales promelas.