差异表达基因鉴别的SAM和RVM的比较  

Identifying Differential Expression Genes with Methods of SAM and RVM

作　　者：贺宪民[1] 武建虎[1] 贺佳[1] XIANG Zhaoying 

机构地区：[1]第二军医大学卫生统计学教研室,200433

出　　处：《中国卫生统计》2005年第4期210-213,共4页Chinese Journal of Health Statistics

基　　金：第二军医大学青年基金资助项目(2003SQ19)

摘　　要：目的探讨小样本量情况下基于两种不同理论对方差进行调整的SAM和RVM方法在差异表达基因鉴别中的性能。方法以真实实验数据作为参数估计的基础进行理论数据的模拟,并基于理论数据对SAM、基于permutation的SAM和RVM的性能进行了探讨。结果基因的对数表达比的残差方差较好地符合反伽马分布,在设定的不同差异表达基因比例情况下,基于permutation的SAM法的鉴别性能最高,RVM次之,SAM最低。结论基因残差方差的分布可以用反伽马分布拟合,基于permutation的SAM和RVM具有较好的检测差异表达基因的性能,SAM在估计校正系数时应考虑差异表达基因的比例。Objective To explore the performance of RVM and SAM in identifying differential expression genes in little sample condition.Methods Simulating theoretical distribution data on the base of actual experimental results, which is used in the analysis of RVM, SAM and SAM based on permutation. Results Inverse-Gamma distribution models true variance structure of microarray data surprisingly well; At different proportions of differential expression genes defined beforehand, the performance of SAM based on permutation is the best, and then RVM and SAM. Conclusion Residual variance of microarray data can be fitted using InverseGamma distribution;The RVM and SAM based on permutation behave high performance in identification of differential expression genes;The proportion of differential expression genes should be considered when estimate the adjust coefficient of SAM.

关 键 词：差异表达基因 随机方差模型 SAM 比较分析 卫生统计 

分 类 号：Q786[生物学—分子生物学]