Statistical Design & Analysis of Microarray Experiments

描述（由申请人提供）：DNA微阵列是一种新的有前途的生物技术，可以同时监测数千个基因的表达水平。微阵列越来越多地应用于生物和医学研究，以解决各种各样的问题，例如肿瘤的分类和宿主基因组对细菌感染的反应的研究。这个项目的长远目标是为设计和分析DNA微阵列实验开发新的统计方法。该提案的具体目标分为四个领域，所有这些领域都涉及提高微阵列实验的效率和可靠性，从早期设计和预处理阶段到更高层次的分析。实验设计。适当的实验设计对于确保在实验限制条件下准确回答生物学问题至关重要。灵活的设计和分析方法将用于时间序列和多因素实验，监测基因表达随时间的反应，如治疗和细胞类型。2。预处理。图像分析和归一化是所有微阵列实验的组成部分，可以对更高水平的分析产生重大影响。将导出斑点和载玻片质量统计数据以及将这些统计数据纳入后续分析的程序。提出了基于鲁棒局部回归的归一化方法，以适应不同类型的染料偏差，并利用在阵列上发现的控制序列。病了。模式发现和识别。为了阐明基因表达水平与其他协变量或反应之间的关系，提出了聚类、鉴别和多重检验的新方法。这包括评估治疗干预措施的效果，发现时间或空间基因表达模式，以及识别与临床结果（如癌症发病率和生存率）相关的基因。四、软件开发。作为该项目的一部分开发的统计方法将在基于R语言构建的统计计算包中实现。为了方便使用和整合生物信息资源，将提供一个网页浏览器界面。

DESCRIPTION (provided by applicant): DNA microarrays are a new and promising biotechnology which allow the monitoring of expression levels for thousands of genes simultaneously. Microarrays are being applied increasingly in biological and medical research to address a wide range of problems, such as the classification of tumors and the study of host genomic responses to bacterial infections. The broad, long-term objectives of this project are to develop novel statistical methods for the design and analysis of DNA microarray experiments. The specific aims of the proposal fall into four areas, all of which are concerned with improving the efficiency and reliability of microarray experiments, from the early design and pre-processing stages to higher level analyses.I. Experimental design. Proper experimental design is essential to ensure that biological questions are answered accurately and precisely given experimental constraints. Flexible designs and methods of analysis will be developed for time-series and multifactorial experiments, which monitor the gene expression response over time for factors such as treatment and cell type. II. Pre-processing. Image analysis and normalization are components of all microarray experiments and can have a substantial impact on higher level analyses. Spot and slide quality statistics will be derived as well as procedures for incorporating these statistics in subsequent analyses. Normalization methods based on robust local regression are proposed to accommodate different types of dye biases and to exploit control sequences spotted on the array. Ill. Pattern discovery and recognition. New methods for clustering, discrimination, and multiple testing are proposed in order to elucidate associations between gene expression levels and other covariates or responses. This includes assessing the effects of treatment interventions, the discovery of temporal or spatial gene expression patterns, and the identification of genes associated with clinical outcomes such as cancer incidence and survival. IV. Software development. Statistical methods developed as part of this project will be implemented in packages built on the R language for statistical computing. To facilitate use and integration with biological information resources a web-browser interface will be provided.