微阵列数据分析和错误发现率

Microarray data analysis and false discovery rate

目的:介绍微阵列数据的差异表达分析和基于错误发现率的多重假设检验。方法:通过t检验对一个关于前列腺癌的微阵列数据进行基因差异表达分析,采用BH程序进行错误发现率的控制和经验估计。结果:当错误发现率为0.05时通过BH程序得到21个差异表达基因;当以|t|≥3作为拒绝域时,得到105个基因,对应的错误发现率估计值为0.20。结论:相对传统的总体错误率,错误发现率更加适合于微阵列这种高维数据多重比较的错误控制;而且能同时控制或估计错误发现率。

Aim：To introduce the analysis of differential expression of microarray data and the multiple hypotheses testing based on the false discovery rate（FDR）.Methods：The t test was used for the analysis of differentially expressed genes concerning prostate cancer microarray data.FDR controlled with the procedure of Benjamini and Hochberg（BH）was empirically estimated.Results：A total of 21 differentially expressed genes were obtained by the BH procedure with the FDR of 0.05;and 105 genes were obtained with an estimated FDR of 0.20 if the rejection region was ｜t｜≥3.Conclusion：FDR is more appropriate for high-dimensional microarray data in multiple comparisons than family wise error rate;we can control and estimate the FDR at the same time.