Statistical Methods to Interpret Genomic Data

最终目标是开发和应用计算和统计技术，对最近出现的基因组数据进行大规模分析，以提取最佳和有意义的生物信息。实现这一目标的具体目标有三个。(1)深入了解基因作图和微阵列实验的技术和生物学方面，并确定功能基因组研究中涉及的统计学问题。这一学习阶段将使PI这位数理统计学家熟悉人类遗传学的生物学。(2)建立单核苷酸多态(SNPs)与特定遗传病的关联分析方法。对于许多复杂的性状，研究中的家系或受影响的个体的数量比基因组筛查中使用的SNP标记的数量要少(或不多)。这排除了在全基因组基础上进行有意义的多变量分析的可能性。为了选择合适的标记子集进行进一步研究，并在多变量分析中结合多个标记的信息，将开发新的统计自举(基于重抽样)方法，所得到的SNP标记子集将进一步通过Logistic或其他多元回归模型进行风险评估。(3)开发基于微阵列技术的基因表达数据的统计分析方法。多个实验的重复性和信号功能等问题经常混淆微阵列数据的分析。提出了一种基于寡核苷酸表达阵列原始数据的多步骤程序，并将开发计算机程序。培训期间开发的方法最终将改进对基因组和基因表达阵列数据的分析。

The ultimate goal is to develop and apply computational and statistical techniques for large-scale analysis of recently emerged genomic data in order to extract optimal and meaningful biological information Specific aims to reach this goal are threefold. (1) To thoroughly understand the technological and biological aspects of gene mapping and microarray- based experiments and to identify the statistical problems involved in functional genomic studies. This learning phase will make the PI, a mathematical statistician, familiar with the biology of human genetics. (2) To develop methods for efficient association analysis between Single Nucleotide Polymorphisms (SNPs) and specific heritable diseases. For many complex traits, the number of families or affected individuals in a study is smaller (or not much larger) than the number of SNP markers used in a genomic screen. This precludes meaningful multi-variate analyses on a genome-wide basis. To selective appropriate subsets of markers for further study and to combine information over multiple markers in multi-variate analyses, novel statistical bootstrap (resampling-based) methods will be developed The resulting subset of SNP markers will be further evaluated by logistic or other multiple regression models for risk assessment. (3) To develop statistical analysis methods for gene expression data obtained through microarray-based technologies. Issues such as reproducibility in multiple experiments and signal function frequently confound the analysis of microarray data. A multi-step procedure based on raw data from oligonucleotide expression array is proposed and computer programs will be developed. The approaches developed in the training period will ultimately allow improved analysis of both genomic and gene expression array data.