Integrative Analysis of Cross-Platform Microarray Data

描述（由申请人提供）：微阵列基因表达谱分析在许多实验室进行，导致公共存储库中的数据快速积累。然而，由于存在不同的技术平台和缺乏标准的实验方案，数据集之间的系统性差异往往超出了统计归一化的能力。目前，迫切需要一种集成跨平台微阵列数据的方法。本提案解决了这一需求。我们的目标是开发新的计算和统计方法来整合跨平台微阵列数据。具体来说，我们将(1)在许多微阵列数据集中检测重复表达模式；(2)对多个基因组进行功能和转录注释；(3)在没有蛋白质- dna结合位点先验信息的情况下预测高级真核基因的转录调控因子；(4)识别疾病特征的基因网络。使用我们的方法，我们能够为任何基因组提取一个数量级的信息，因为大量的微阵列数据是可用的。我们将对酵母（S. cerevisiae）、蠕虫（C. elegans）、苍蝇（D. melanogaster）、植物（A. thaliana）、小鼠（M. musculus）、大鼠（R. norvegicus）和人类（H. sapiens）的基因组进行“上下文特异性”功能和转录注释。也就是说，我们将有条件地注释基因的功能/调控，这取决于它们与哪组其他基因相互作用以及在哪些条件下发生这种相互作用。在发布预测结果时，我们将为每个注释附加必要的上下文信息。最后，我们将开发一个软件包ARRAYMINE，供生物学家进行跨平台微阵列数据的综合分析。我们的算法和软件将极大地促进大量现有微阵列数据的再利用，减少生成新数据的必要性，并提高我们对各种扰动下细胞功能和网络的理解。

DESCRIPTION (provided by applicant): Microarray gene expression profiling is performed in many laboratories, resulting in the rapid data accumulation in public repositories. However, due to the existence of different technology platforms and the lack of standard experimental protocols, systematic variation among data sets often exceeds the capability of statistical normalization. Currently, there is an urgent need for methodology to integrate cross-platform microarray data. This proposal addresses this need. We aim at developing novel computational and statistical methods to integrate cross-platform microarray data. Specifically, we will (1) detect recurrent expression patterns across many microarray datasets; (2) perform functional and transcriptional annotation for multiple genomes; (3) predict transcription regulators for higher eukaryotic genes without prior information on protein-DNA binding sites; and (4) identify genetic networks that are signatures of diseases. Using our approach, we are in a position to extract an order of magnitude more information for any genome for which massive microarray data is available. We will perform "context-specific" functional and transcriptional annotation for the genomes of yeast (S. cerevisiae), worm (C. elegans), fly (D. melanogaster), plant (A. thaliana), mouse (M. musculus), rat (R. norvegicus) and human (H. sapiens). That is, we will conditionally annotate the functions/regulations of genes, depending on which set of other genes they are interacting with and under which sets of conditions such interactions occur. When releasing our prediction results, we will attach to each annotation the necessary context information. Finally, we will develop a software package ARRAYMINE for biologists to perform integrative analysis of cross-platform microarray data. Our algorithms and software will significantly facilitate the re-use of the vast amount of existing microarray data, reduce the necessity to generate new data, and improve our understanding of cellular functions and networks under a variety of perturbations.