论文部分内容阅读
在新药设计中需要根据化合物的化学结构参数的差别进行分类。有时也需要判断某一化合物属于何种类别。如单纯依靠经验来归纳,或仅凭直观来判断,往往只能定性地分类,粗略地判定归属问题。但运用多元统计分析方法中的判别分析(Discriminant analysis),则分类和判断归属既能定量表示,又较客观。因此,判别分析为数学分类的一种多元统计分析方法。既可在已知化合物生物活性基础上,根据化合物的多个化学结构参数(变量),总结出分类的规律性(即判别公式),进行合理分类。又可在分类规律性基础上,判断新化合物属于何类,以估计其生理活性。
In the design of new drugs need to be classified according to the chemical structure of the compound parameters. Sometimes you need to determine which compound a compound belongs to. Such as simply relying on experience to summarize, or just to judge by intuition, often only qualitatively classified, roughly determine the ownership problem. However, using discriminant analysis in multivariate statistical analysis, classification and judgment of attribution can be both quantitative and objective. Therefore, discriminant analysis is a multivariate statistical analysis method of mathematical classification. Based on the biological activity of the known compounds, the regularity of the classification (ie, the discriminant formula) can be summed up based on the chemical structural parameters (variables) of the compounds and the reasonable classification can be made. Again on the basis of the classification of the law, to determine what the new compounds belong to in order to assess their physiological activity.