Freitas Renato F, Oprea Tudor I, Montanari Carlos A
Grupo de Química Medicinal de Produtos Naturais, NEQUIMED-PN, Departamento de Química e Física Molecular, Instituto de Química de São Carlos, Universidade de São Paulo, Av. Trabalhador Sancarlense, 400, 13560-970 São Carlos, SP, Brazil.
Bioorg Med Chem. 2008 Jan 15;16(2):838-53. doi: 10.1016/j.bmc.2007.10.048. Epub 2007 Oct 22.
Hologram quantitative structure-activity relationships (HQSAR) were applied to a data set of 41 cruzain inhibitors. The best HQSAR model (Q(2)=0.77; R(2)=0.90) employing Surflex-Sim, as training and test sets generator, was obtained using atoms, bonds, and connections as fragment distinctions and 4-7 as fragment size. This model was then used to predict the potencies of 12 test set compounds, giving satisfactory predictive R(2) value of 0.88. The contribution maps obtained from the best HQSAR model are in agreement with the biological activities of the study compounds. The Trypanosoma cruzi cruzain shares high similarity with the mammalian homolog cathepsin L. The selectivity toward cruzain was checked by a database of 123 compounds, which corresponds to the 41 cruzain inhibitors used in the HQSAR model development plus 82 cathepsin L inhibitors. We screened these compounds by ROCS (Rapid Overlay of Chemical Structures), a Gaussian-shape volume overlap filter that can rapidly identify shapes that match the query molecule. Remarkably, ROCS was able to rank the first 37 hits as being only cruzain inhibitors. In addition, the area under the curve (AUC) obtained with ROCS was 0.96, indicating that the method was very efficient to distinguishing between cruzain and cathepsin L inhibitors.
全息定量构效关系（HQSAR）被应用于41种克氏锥虫蛋白酶抑制剂的数据集。使用Surflex-Sim作为训练集和测试集生成器，以原子、键和连接作为片段区分，片段大小为4 - 7，获得了最佳的HQSAR模型（Q(2)=0.77；R(2)=0.90）。然后使用该模型预测12种测试集化合物的效力，得到令人满意的预测R(2)值0.88。从最佳HQSAR模型获得的贡献图与研究化合物的生物活性一致。克氏锥虫的克氏锥虫蛋白酶与哺乳动物同源物组织蛋白酶L具有高度相似性。通过一个包含123种化合物的数据库检查对克氏锥虫蛋白酶的选择性，该数据库对应于HQSAR模型开发中使用的41种克氏锥虫蛋白酶抑制剂加上82种组织蛋白酶L抑制剂。我们通过ROCS（化学结构快速叠加）筛选这些化合物，ROCS是一种高斯形状体积重叠过滤器，能够快速识别与查询分子匹配的形状。值得注意的是，ROCS能够将前37个命中结果列为仅为克氏锥虫蛋白酶抑制剂。此外，ROCS获得的曲线下面积（AUC）为0.96，表明该方法在区分克氏锥虫蛋白酶抑制剂和组织蛋白酶L抑制剂方面非常有效。
