Instituto de Investigaciones Fisicoquímicas Teóricas y Aplicadas, UNLP, CCT La Plata-CONICET, Diag. 113 y 64, C.C. 16, Suc.4, 1900 La Plata, Argentina.
Eur J Med Chem. 2010 Nov;45(11):4879-89. doi: 10.1016/j.ejmech.2010.07.060. Epub 2010 Aug 7.
Leishmaniasis is a neglected disease transmitted in many tropical and sub-tropical countries, with few studies devoted to its treatment. In this work, the activities of two antileishmanial compound classes were modeled using Dragon descriptors, and multiple linear (MLR) and support vector machines (SVM) as linear and nonlinear regression methods, respectively. Both models were highly predictive, with calibration, leave-one-out validation and external validation R(2) of 0.79, 0.72 and 0.78, respectively, for the MLR-based model, improving significantly to 0.98, 0.93 and 0.90 when using SVM modeling. Therefore, novel compounds were proposed using the QSAR models built by combining the substructures of the main active compounds of both classes. The most promising structures were docked into the active site of Leishmania donovani α,β tubulin (Ld-Tub), demonstrating the high affinity of some new structures when compared to existing antileishmanial compounds.
利什曼病是一种在许多热带和亚热带国家传播的被忽视的疾病，针对其治疗方法的研究很少。在这项工作中，使用 Dragon 描述符对两种抗利什曼病化合物进行了活性建模，多元线性回归（MLR）和支持向量机（SVM）分别作为线性和非线性回归方法。这两种模型都具有很高的预测能力，基于 MLR 的模型的校准、留一法验证和外部验证 R(2)分别为 0.79、0.72 和 0.78，而使用 SVM 建模时，分别显著提高到 0.98、0.93 和 0.90。因此，使用两种化合物类别的主要活性化合物的亚结构组合构建 QSAR 模型，提出了新的化合物。最有前途的结构被对接进入利什曼原虫 α，β 微管蛋白（Ld-Tub）的活性位点，与现有的抗利什曼病化合物相比，一些新结构表现出了很高的亲和力。
