目的 中药是一个具有复杂组分的统一体,无论是方剂还是单味中药,其药效都是其中多种化学成分相互作用的综合效果,具有多组分、多靶点、多渠道作用的特点.本文阐述利用代谢组学方法研究中药作用机理的数据分析方法及特点,为医学工作者提供新的中药研究思路及策略.方法 从生物统计学和生物信息学角度,利用文献和结合目前相关研究结果提出作者的观点和看法.代谢组数据分析的主要困难是相对于给定的样品数目谱峰的数量巨大,用传统的统计方法对可能具有生物学意义的"差异谱峰"进行鉴别会产生大量的假阳性结果.特征选择方法按照算法可分为过滤法、包裹法和嵌入法,三种方法各有特点.结果 代谢组指纹图谱数据能够为我们提供大量药物化学、特别是次生代谢物质的信息,对这种高维数据有多种分析方法可以使用,如果不对数据做变量筛选,难免受大量对分类不起作用的无关变量的干扰.变量筛选有很多优点:既可以简化模型,利于可视化和数据解释,同时可以更好地避免维数灾难引起的过拟合问题,提高模型分类效果.代谢组数据库和一些软件也是我们可以利用的工具.结论 利用代谢组学的方法研究中药的作用机理是一种可行的方法,研究中药的代谢指纹图谱应包括化学和药效两方面的的内容,为有效提取其生物学信息,必须采取适宜的统计学模型结合生物学知识对其进行研究.
Objective Traditional Chinese Medicine (TCM) is a complex prescription of a combination of several components. Prescriptions or herbs, the efficacy is a combined effect of multiple chemical components with characteristics of multicomponents, multi-target sites and multi-channels. This paper studies the methods and characteristics of data analysis of TCM mechanism with metabonomic data analysis to provide medical researchers with the new idea and strategy of the study of TCM mechanism. Methods The views and perspectives were presented in this paper based on the literatures and related research findings in the biostatistics and bioinformatics areas. Currently, the main difficulty in metabonomie data analysis lies in the im- mense number of GC-MS peaks in terms of given samples and the traditional statistical analysis methods may generate a large number of false positive results for the differential peaks of biological significance. In generally, the component selection methods include the infiltration, package and embedding techniques and these methods have their respective strengths and drawbacks. Re- suits Metabonomics fingerprinting can provide sufficient information about medicinal chemistry, especially about its secondary metabolites and many statistical analysis methods for high-dimensional metabonomic data are available. If component selection is not performed to deal with metabonomie data prior to analysis, the results may be influenced by irrelevant variables. There are a large number of advantages for metabolic components selection, such as simplifying the model, visualizing and avoiding over-fitting. In addition, the metabonomie datasets and application software are also useful for the metabonoie data analysis. Conclusion It is a feasible approach to investigate the mechanism of TCM with metabonomie technology. Because the metabolic fingerprinting consists of the chemistry and efficacy contents, appropriate statistical models in combination with biological knowledge is needed to extract biological i