Bitencourt Michelle, Freitas Matheus P
Departamento de Química, Universidade Federal de Lavras, CP 3037, 37200-000 Lavras, MG, Brazil.
Pest Manag Sci. 2008 Aug;64(8):800-7. doi: 10.1002/ps.1565.
A series of sulfonylurea herbicides has been modelled using a 2D image-based QSAR approach known as MIA-QSAR (Multivariate Image Analysis applied to QSAR), and highly predictive models have been built.
Two MIA-QSAR models were built, one group being divided into training and test sets, and the other composed of the entire series of compounds. Statistically significant MIA-QSAR models rendered high correlation coefficients of experimental versus fitted pK(i)(app) (AHAS apparent inhibition constant) and satisfactory parameters of external validation and leave-one-out cross-validation. Comparison with the results obtained from classical 2D QSAR demonstrated some advantages of the modelling using MIA descriptors.
Both MIA-QSAR models showed high predictive ability, comparable with that of a reference methodology based on 3D descriptors. The method is suggested as a suitable tool for predicting novel herbicides.
已使用一种名为MIA-QSAR（应用于定量构效关系的多变量图像分析）的基于二维图像的定量构效关系方法对一系列磺酰脲类除草剂进行建模，并建立了高度预测性模型。
建立了两个MIA-QSAR模型，一组分为训练集和测试集，另一组由整个化合物系列组成。具有统计学意义的MIA-QSAR模型给出了实验值与拟合pK(i)(app)（AHAS表观抑制常数）的高相关系数，以及令人满意的外部验证和留一法交叉验证参数。与经典二维定量构效关系获得的结果比较表明，使用MIA描述符进行建模具有一些优势。
两个MIA-QSAR模型均显示出高预测能力，与基于三维描述符的参考方法相当。该方法被认为是预测新型除草剂的合适工具。
