几种差异基因分析方法及筛选效果的比较  

Various Statistical Methods for Identifying Differentially Expressed Genes and a Comparative Study Using New Methods

作　　者：赵发林[1] 闫晓光[1] 李康[1] 

机构地区：[1]哈尔滨医科大学卫生统计学教研室,150081

出　　处：《中国卫生统计》2008年第4期354-356,362,共4页Chinese Journal of Health Statistics

基　　金：国家自然科学基金资助(30371253);黑龙江省重点项目(GB04C30202)

摘　　要：目的比较六种差异基因筛选方法的使用效果及适用性。方法用Monte-Carlo方法产生不同类型的模拟数据,分别用不同的方法计算、评价其优劣。结果多数情况下SAM法和稳健t检验表现出了最优的筛选能力,SAM-ROC法则表现出更好的稳定性。结论几种方法都能够有效地用于基因筛选,但各自的适应条件不同,综合看SAM法是基因筛选的首选方法,随机森林方法则具有较大的研究价值。Objective To explore the characteristics of six statistical methods in identifying differentially expressed genes by comparing their performances in various conditions. Methods We applied these methods to the data simulated with Monte-Carlo method in various condi tions . Results In the most conditions, SAM and moderated t statistic perform best capability of selecting differential expression genes, while SAMROC is the stablest one and the capability of selecting differential gene is good, too. Conclusion We prefer SAM to other methods in general, but there suitable situations are slightly different from each other. The random forest may answer to the biological characteristics of microarray data better.

关 键 词：微阵列数据 基因筛选 Monte—Carlo模拟 

分 类 号：Q3[生物学—遗传学]