【摘要】 利用SAS软件对GEO的一个肺癌芯片实验进行挖掘。采用非参数检验,判别分析和回归分析对该芯片实验中14个核受体的表达信息进行分析。结果表明,在0.05显著性水平下,ER1、VDR、RARα和RORα四个基因在腺癌和鳞癌表达具有统计学差异;RARβ在复发组和非复发组表达有差异。判别分析结果显示VDR和RORα表达量可以对病理类型进行预测,但是总误判率很高(0.2389);RARβ和PPARα对判别是否复发的总误判率更高(0.3457)。建立回归方程预测病理类型,入选模型的变量也是VDR和RORα,两者OR分别为0.126和4.452。可见,基于SAS的多元统计方法是芯片数据挖掘的一种潜在方法,一旦芯片实验标准化,利用SAS对不同芯片实验数据整合分析的结论将有益于推动假说形成。更多还原

【Abstract】 Multivariate statistics using SAS is applied to mine a dataset from GEO.Expression data of fourteen nuclear receptors in a lung cancer microarray experiment is analyzed by non-parameter test,discriminant analysis and regression analysis.As a result,ER1,VDR,RARα and RORα is differentially expressed between adenocarcinoma and squamous cell carcinoma under significance of 0.05;RARβ is differentially expressed between recurrent and non-recurrent cancer;discriminant analysis shows VDR and RORα together can predict pathotype,and RARβ and PPARα together can discriminate recurrence;the false-rate is 0.2389 and 0.3457,respectively.Logistic regression is established to predict pathotype and variables included are also VDR and RORα,with OR at 0.126 and 4.452,respectively.Therefore,multivariate statistics based on SAS is a potential way to mine microarray data and conclusions based on SAS integration of different microarray experiments might be helpful for establishing hypothesis once microarray experiments can be standardized.更多还原