分子电性-距离矢量对典型抗HIV类药物的有效表征

QSAR of typical anti-HIV molecular based on MEDV descriptors

采用二维拓扑结构描述子—分子电性距离矢量(MEDV),对TIBO类衍生物、非核苷类逆转录酶抑制剂和环脲类化合物等三类抗HIV病毒的药物定量结构活性关系进行了研究.用MEDV描述子对三类药物分子进行结构表征,以VSMP法对所得的91个分子结构描述符进行优化筛选以得到最优子集,用多元线性回归方法(MLR)分别构建了三类抗HIV病毒药分子结构与生物活性之间的定量构效关系(QSAR)模型.模型相关系数r分别为0.930 4,0.852 5和0.957 9,留一法交互验证(leave-one-out cross validation,LOO-CV)相关系数q分别为0.888 8,0.815 5和0.908 4,表明QSAR模型具有良好的稳定性和预测能力.QSAR模型表明分子中取代基团的种类和取代位置对药物分子的生物活性有显著影响.

Using a kind of 2-D topological descriptors which is called Molecular Electronegativity-Distance Vector （MEDV）, some quantitative structure-activity relationship （QSAR） models related the MEDV descriptors to the bioactivities of three sets of anti-H[V pharmacy molecular are developed. The MEDV descriptors are used to characterize and represent the structure of these three sets of compounds, then a VSMP program is used to optimize the obtained 91 variables. By multiple liner regression method, the QSAR models between molecular structures and the biological activity are established. In these three sets of compounds, the regressive correlation coefficient （r） and the correlation coefficient in prediction model （q） for TIBO compounds is 0.9304 and 0.8888 respectively. The resemble correlation coefficient and the correlation coefficient in prediction model of nevirapine derivatives is 0.8525 and 0.8155 respectively. Similarly, the values of r and q of cyclic urea is 0.9579 and 0.9084 respectively. The quantitative structure-activity relationship models based on MEDV vector descriptors have a good stability and predictive ability.