VSMVI方法预测取代芳烃类化合物淡水发光菌的毒性

发布时间：2018-06-02 10:31

本文选题：取代芳烃类化合物 + 定量结构-毒性关系　；参考：《桂林理工大学学报》2016年04期

【摘要】：选取2组取代芳烃类共53个化合物为目标化合物,以Dragon描述符表征其分子结构信息,采用基于变量相互作用的变量筛选方法(VSMVI)对训练集化合物进行变量寻优。以Dragon描述符为自变量,以化合物的毒性值作为因变量,然后根据化合物样本数和奥卡姆剃刀原则,运用通用变量筛选与建模程序(GVSMP)中线性建模的最优建模法,建立了化合物分子结构与其毒性间的相互关系模型——定量结构-毒性关系(QSTR)模型。结果表明,模型的复相关系数R~2和交互验证相关系数Q2LOO均在0.90以上,所构建的模型均具有较高的拟合能力和预测能力。此外,通过模型校验方法对模型进行校验,也显示出模型具有较好的外部估计能力。
[Abstract]:Two groups of 53 substituted aromatic compounds were selected as target compounds and their molecular structure information was characterized by Dragon descriptor. Variable selection method based on variable interaction was used to optimize the training set compounds. Taking the Dragon descriptor as the independent variable and the toxic value of the compound as the dependent variable, according to the sample number of compounds and the Occam razor principle, the optimal modeling method of linear modeling is used in the general variable screening and modeling program (GVSMP). A quantitative structure-toxicity relationship model (QSTR) was established for the relationship between the molecular structure of compounds and their toxicity. The results show that both the complex correlation coefficient Rn2 and the cross-validation correlation coefficient Q2LOO are above 0.90, and the models constructed have higher fitting ability and prediction ability. In addition, the model verification method also shows that the model has good external estimation ability.
【作者单位】：桂林理工大学广西高校食品安全与检测重点实验室;桂林理工大学化学与生物工程学院;
【基金】：国家自然科学基金项目(21467006) 广西自然科学基金项目(桂科自2014GXNSFAA118034) 广西高校食品安全与检测重点实验室项目
【分类号】：X171.5

【相似文献】