Empirical assessment of analysis methods for DNA microarrays

描述（由申请人提供）： DNA微阵列是生物学和生物医学研究学科中越来越重要的技术。随着微阵列的普及，提出了大量分析微阵列数据的方法。然而，评估这些各种方法的有效性一直是困难的，因为"正确"的答案通常是未知的;也就是说，由于在微阵列实验中询问的基因的大量，只有相对小部分的基因表达差异往往在任何给定的研究中得到验证。虽然已经进行了一些尝试来解决这个问题并在不同的分析方法之间进行比较，但它们倾向于使用少量的已知对照基因，使得充分的统计分析变得困难，并且包括未知组成的大背景RNA样品，防止准确评估假阳性率和与阵列的非特异性杂交。我们最近报道了一个新的控制数据集的目的，评估微阵列分析方法，适用于Affytek基因芯片。该数据集具有三个关键特征：它包含超过1300种已知相对浓度不同的RNA;它包括低倍数变化，从1.2 x浓度差异开始，以便可以充分评估微阵列的灵敏度;它包含超过2500种RNA的定义背景样本。使用这个对照数据集，我们能够评估一些流行的分析方法的Affytron阵列，并设计一个新的算法，降低假阴性率。在这里，我们提出了一个类似的，但改进的控制数据集的建设，以应用于斑点载玻片微阵列，这在其分析要求不同，从Affyoung阵列在几个领域，也是一个改进的控制数据集扩展Affyoung研究。 DNA微阵列不仅越来越多地被用作生物医学研究中的基本工具，特别是在癌症和神经退行性疾病领域，而且它们还具有用于异质性疾病（如癌症）的临床测试的重要潜力。 为了实现这些研究和诊断潜力，我们必须知道如何准确地分析微阵列数据。 这项建议涉及这一重要问题。

DESCRIPTION (provided by applicant): DNA microarrays are an increasingly important technology for a diverse spectrum of biological and biomedical research disciplines. Concomitant with the rise in microarrays' popularity has been a surge of proposed methods for the analysis of microarray data. However, assessing the effectiveness of these various methods has been difficult as the "correct" answers typically are not known; that is, due to the vast numbers of genes interrogated in a microarray experiment, only a relatively small fraction of gene expression differences tend to be validated in any given study. While some attempts to address this problem and to compare among different analysis methods have been made, they have tended to use small numbers of known control genes, making adequate statistical analysis difficult, and have included large background RNA samples of unknown composition, preventing an accurate assessment of false positive rates and nonspecific hybridization to the array. We recently reported a new control dataset for purposes of evaluating microarray analysis methods as applied to Affymetrix GeneChips. This dataset has three key features: it contains over 1300 RNAs that differ by known relative concentrations; it includes low fold changes, beginning at a 1.2 x concentration difference, so that sensitivity of the microarrays can be fully assessed; and it contains a defined background sample of over 2500 RNAs. Using this control dataset we were able to assess a number of popular analysis methods for Affymetrix arrays and to devise a new algorithm that reduced false-negative rates. Here, we propose the construction of a similar but improved control dataset to be applied to spotted glass slide microarrays, which differ in their analysis requirements from Affymetrix arrays in several areas, and also an improved control dataset to extend the Affymetrix studies. DNA microarrays are not only increasingly being used as a fundamental tool in biomedical research, particularly in the areas of cancer and neurodegenerative diseases, but they also have important potential for use in clinical testing for heterogeneous diseases such as cancer. In order for these research and diagnostic potentials to be realized, it is vital that we know how to analyze microarray data accurately. This proposal addresses this important issue.