Garcia-Domenech Ramon, Zanni Riccardo, Galvez-Llompart Maria, Galvez Jorge
Department of Physical chemistry, University of Valencia, Avenida V.A. Estelles s/n, 46100, Burjassot, Valencia, Spain,
Mol Divers. 2015 May;19(2):357-66. doi: 10.1007/s11030-015-9575-5. Epub 2015 Mar 10.
Multi-target QSAR is a novel approach that can predict simultaneously the activity of a given chemical compound on different pharmacological targets. In this work, we have used molecular topology and statistical tools such as multilinear regression analysis and artificial neural networks, to achieve a multi-target QSAR model capable to predict the antiprotozoal activity of a group of benzyl phenyl ether diamine derivatives. The activity was related to three parasites with a high prevalence rate in humans: Trypanosoma brucei rhodesiense, Plasmodium falciparum, and Leishmania donovani. The multi-target model showed a high regression coefficient (R(2) = 0.9644 and R(2) = 0.9235 for training and test sets, respectively) and a low standard error of estimate (SEE = 0.279). Model validation was performed with an external test (R(2) = 0.9001) and a randomization analysis. Finally, the model was applied to the search of potential new active compounds.
多靶点定量构效关系（QSAR）是一种新型方法，能够同时预测给定化合物对不同药理靶点的活性。在本研究中，我们运用了分子拓扑学和统计工具，如多元线性回归分析和人工神经网络，以建立一个能够预测一组苄基苯基醚二胺衍生物抗寄生虫活性的多靶点QSAR模型。该活性与人类中流行率较高的三种寄生虫有关：布氏罗得西亚锥虫、恶性疟原虫和杜氏利什曼原虫。多靶点模型显示出较高的回归系数（训练集和测试集的R²分别为0.9644和0.9235）以及较低的估计标准误差（SEE = 0.279）。通过外部测试（R² = 0.9001）和随机化分析进行模型验证。最后，该模型被应用于寻找潜在的新型活性化合物。
