代谢组学研究中样本间区分度的简易评判方法  

Evaluation Methods for Discrimination of Metabonomic Data

作　　者：朱航[1] 蓝文贤[2] 刘买利[1] 

机构地区：[1]波谱与原子分子物理国家重点实验室武汉磁共振中心中国科学院武汉物理与数学研究所,武汉430071 [2]中国科学院上海有机化学研究所,上海200032

出　　处：《分析化学》2013年第7期1000-1005,共6页Chinese Journal of Analytical Chemistry

基　　金：国家自然科学基金(No.20635054)资助项目

摘　　要：在代谢组学现有数据分析方法的基础上,引入了多元统计分析方法中的多元t检验、类间距计算、组内离差平方和计算等统计方法。结合4个代谢组学研究实例,演示了新引入的统计方法的应用。结果表明:多元t检验可以解答应激组和对照组的代谢组水平之间是否存在"系统差别",类间距可以提供差别大小,组内离差平方和可以提供离散程度等信息。这些新信息的获取能对模式识别分析所得出的得分图(Scores plot)作进一步的解释,并为后续实验设计提供参考。Metabonomics, which utilizes spectroscopic approaches to obtain the data and statistic approaches to analyze the data, has been widely used to study metabolic response of living system to internal and external stimuli. In this study, multivariate t test, interclass-distance and sum of squares of deviations were introduced for better evaluating the metabonomic data, aiming to answer the questions such as ＂Is there significant diffe- rence between classes＂, ＂How big is the difference＂ and ＂How large is the data disparity＂ in a quantitative or semi-quantitative manner. The results demonstrated that the introduced statistic approaches could provide more information for the data and might be useful in designing experiment for metabolic studies.

关 键 词：代谢组学 模式识别 多元统计分析 显著性检验 类间距 离差平方和 核磁共振 

分 类 号：R341[医药卫生—基础医学]