本文从氨基酸的结构特征出发,构建了氨基酸结构表征体系,即P-scale(principal component score vector of hydrophobic,electronic,steric,and hydrogen bond physicochemical properties)来源于20种天然氨基酸的113种物理化学性质,通过对其中50个疏水性质,23个电性性质,35个立体性质和5个氢键性质分别进行主成分特征提取而产生。并将P-scale描述子应用于101个阳离子抗菌肽QSAR研究。结果表明,基于遗传算法-偏最小二乘法的QSAR模型具有很好的预测能力,发现抗菌肽序列信息与抗菌效价有很强相关性,其中,阳离子抗菌肽第3残基电性性质,第6、7和12残基立体性质以及第11和12残基的疏水性与抗菌效价呈正相关关系,而第6、10和12残基电性性质则与抗菌效价呈显著的负相关。基于良好QSAR模型,我们设计出高活性抗菌肽序列,为设计分子的合成与活性实验提供借鉴。
The molecular docking by LigandFit docking of Discovery Studios 2.5 was employed to the three-dimensional quantitative structure-activity relationship（3D-QSAR） studies of biphenyl carboxylic acid MMP3 inhibitors.A significant correlation coefficient was obtained between dock scores and biological activities.Based on the optimal docking conformations,3D-HoVAIF was employed to the QSAR studies of 51 biphenyl carboxylic acid MMP-3 inhibitors.R2 and Q_CV2（leave-one-out,LOO） of the optimal 3D-HoVAIF-PLS model were 0.873 and 0.841 respectively.The conclusions obtained from the PLS analysis were in agreement with the docking results.
