Statistical Methods for Normalizing Microarrays in Cancer Biomarker Studies

描述（由申请人提供）：基因表达微阵列已经开发了各种阵列归一化方法。这些方法大多假设样本组之间的微分表达式很少或对称。目前还没有系统研究这些方法在利用异质样本（如肿瘤）规范化microRNA表达阵列中的特性。MicroRNA阵列仅包含几百个MicroRNA，并且在不同肿瘤组之间可能有相对较大的比例存在差异表达。由于缺乏没有混杂数组效应的基准数据集，在这种情况下评估规范化方法是困难的。我们建议设计和生成这样的基准数据集，对规范化方法进行系统评估，特别强调这些模型在检测差异表达标记物方面的效用，并从基准数据设计中推导出承认肿瘤样本固有异质性的统计模型。

DESCRIPTION (provided by applicant): Various array normalization methods have been developed for gene expression microarrays. Most of these methods assume few or symmetric differential expression between sample groups. There has been no systematic study of the properties of these methods in normalizing microRNA expression arrays utilizing heterogeneous samples such as tumors. MicroRNA arrays contain only a few hundred microRNAs, and are likely to have a relatively large proportion being differentially expressed between diverse tumor groups. The assessment of normalization methods in this setting is difficult because of the lack of a benchmark dataset that has no confounding array effects. We propose to design and generate such benchmark datasets, perform a systematic assessment of normalization methods with a particular emphasis on the utility of these models for detecting markers with differential expression, and from the benchmark data design derive statistical models that acknowledge heterogeneities inherent to tumor samples. PUBLIC HEALTH RELEVANCE: Microarrays are being widely used in cancer research. A critical step for processing microarray data is to normalize the arrays so that measurements from different arrays are comparable. There is a great need to evaluate the properties of statistical methods for array normalization when they are applied to microRNA arrays utilizing heterogeneous samples such as tumors.