Rasulev Bakhtiyor, Kusić Hrvoje, Leszczynska Danuta, Leszczynski Jerzy, Koprivanac Natalija
Interdisciplinary Center for Nanotoxicity, Jackson State University, 1400 J.R. Lynch Street, Jackson, Mississippi 39217, USA.
J Environ Monit. 2010 May;12(5):1037-44. doi: 10.1039/b919489d.
The goal of the study was to predict toxicity in vivo caused by aromatic compounds structured with a single benzene ring and the presence or absence of different substituent groups such as hydroxyl-, nitro-, amino-, methyl-, methoxy-, etc., by using QSAR/QSPR tools. A Genetic Algorithm and multiple regression analysis were applied to select the descriptors and to generate the correlation models. The most predictive model is shown to be the 3-variable model which also has a good ratio of the number of descriptors and their predictive ability to avoid overfitting. The main contributions to the toxicity were shown to be the polarizability weighted MATS2p and the number of certain groups C-026 descriptors. The GA-MLRA approach showed good results in this study, which allows the building of a simple, interpretable and transparent model that can be used for future studies of predicting toxicity of organic compounds to mammals.
该研究的目标是通过使用定量构效关系/定量结构-性质关系（QSAR/QSPR）工具，预测由具有单个苯环结构以及存在或不存在不同取代基（如羟基、硝基、氨基、甲基、甲氧基等）的芳香化合物在体内引起的毒性。应用遗传算法和多元回归分析来选择描述符并生成相关模型。最具预测性的模型是三变量模型，该模型在描述符数量与其预测能力之间也具有良好的比例，以避免过度拟合。对毒性的主要贡献显示为极化率加权的MATS2p和某些基团C-026描述符的数量。遗传算法-多元线性回归分析（GA-MLRA）方法在本研究中显示出良好的结果，这使得能够构建一个简单、可解释且透明的模型，可用于未来预测有机化合物对哺乳动物毒性的研究。
