论文部分内容阅读
目的:研究N-(3-氧-3,4-二氢-2H-苯并[1,4](口恶)嗪-6-羰基)胍类化合物的量子化学结构参数与Na/H交换抑制活性的关系,探讨计算机辅助识别该类化合物生物活性的手段。方法:用我们独创的超多面体模型。结果:用超多面体模型总结出该类化合物(样本)在多维空间中的分布区域,建立了该类化合物的构效关系。并用“留一法”比较了PCA、Fisher方法与超多面体模型的预报能力。结论:对同一套数据的胍类化合物分子,用超多面体模型作分子筛选的准确率比用PCA和Fisher方法高。因此,超多面体模型可望成为筛选新高活性化合物的一个有力工具。
Aims: To investigate the relationship between the quantum chemical structure parameters and Na / H exchange inhibition of N- (3-oxo-3,4-dihydro-2H-benzo [1,4] Activity of the relationship between computer-assisted identification of biological activity of these means. Method: Use our original super-polyhedron model. Results: The hyper-polyhedron model was used to summarize the distribution of these compounds in multidimensional space. The structure-activity relationship of these compounds was established. And compared the predictive ability of PCA, Fisher method and hyper-polyhedron model with “leave one method”. Conclusion: For the same set of data of guanidine compounds, the accuracy of molecular screening by hyperdimensional model is higher than that by PCA and Fisher. Therefore, the hyper-polyhedron model is expected to be a powerful tool for screening new highly active compounds.