de Melo Eduardo Borges, Ferreira Márcia Miguel Castro
Centro de Ciências Médicas e Farmacêuticas, Universidade Estadual do Oeste do Paraná, Rua Universitária 2069, 85819-110 Cascavel, PR, Brazil.
Eur J Med Chem. 2009 Sep;44(9):3577-83. doi: 10.1016/j.ejmech.2009.03.001. Epub 2009 Mar 9.
A multivariate QSAR study of thirty-three 4,5-dihydroxypyrimidine carboxamides as HIV-1 integrase (HIV-1 IN) inhibitors was performed employing Ordered Predictors Selection (OPS) algorithm and PLS regression for variable selection and model construction, respectively. Four descriptors were chosen and a reasonable model (n=30; R(2)=0.68; SEC=0.57; PRESS(cal)=8.72; F((2,27))=28.97; Q(2)(LOO)=0.58; SEV=0.62; PRESS(val)=11.62; R(2)(pred)=0.87; SEP=0.29; ARE(pred)=4.37%; k=0.99; k'=1.01; |r(2)(0)-r(2)(0)'|=-0.18) was built with two latent variables (59.54% of the information). Leave-N-out (LNO) and Y-randomization methods confirmed the model robustness. The descriptors indicated that the HIV-1 IN inhibition depends on the electronic distribution of the investigated compounds. The interpretation of the model is related to the most accepted mechanism of action.
采用有序预测变量选择（OPS）算法和偏最小二乘（PLS）回归分别进行变量选择和模型构建，对33种4,5-二羟基嘧啶羧酰胺作为HIV-1整合酶（HIV-1 IN）抑制剂进行了多变量定量构效关系（QSAR）研究。选择了四个描述符，并建立了一个合理的模型（n = 30；R(2)=0.68；SEC = 0.57；PRESS(cal)=8.72；F((2,27))=28.97；Q(2)(LOO)=0.58；SEV = 0.62；PRESS(val)=11.62；R(2)(pred)=0.87；SEP = 0.29；ARE(pred)=4.37%；k = 0.99；k' = 1.01；|r(2)(0)-r(2)(0)'|=-0.18），该模型包含两个潜变量（占信息的59.54%）。留一法（LNO）和Y随机化方法证实了模型的稳健性。这些描述符表明，HIV-1 IN抑制作用取决于所研究化合物的电子分布。该模型的解释与最被认可的作用机制相关。
