Statistical methods for high-throughput genomics

高通量基因组技术（如微阵列和测序技术）使基因表达及其调控机制的全基因组评估成为可能。近年来，这些技术已经彻底改变了生物学和医学领域，但仍然迫切需要开发有效和严格的统计方法来分析由此产生的庞大而复杂的数据集。由于新的测序数据本质上是离散的，我们首先开发了适用于离散数据的强大的多重测试方法。第二个问题出现在测试多组基因在不同条件下的差异表达时，这是高通量基因组数据分析的常见做法。因为基因集可以共享基因，所以测试结果应该满足一定的逻辑约束。我们将开发计算高效的统计方法，用于这些逻辑约束假设的多重检验。随着测序技术成为研究全基因组调控事件的主要工具，开发用于误差率估计和基因组事件检测控制的统计方法变得越来越重要。最后，随着越来越多的高通量基因组数据的积累，我们将开发多数据集联合分析的方法，以提高基因组发现的敏感性和特异性。

High-throughput genomics technologies (e.g. microarray and sequencing technology) enable genome-wide evaluation of genes expression and their regulatory mechanism. Such technologies have revolutionized the fields of biology and medicine in recent years, but there remains an urgent need for the development of efficient and rigorous statistical methods to analyze the resultant large and complex datasets. As the new sequencing data are inherently discrete, we first develop powerful multiple testing methods that are suitable for discrete data. A second problem arises when multiple sets of genes are tested for differential expression across conditions, which is a common practice in high-throughput genomic data analysis. Because gene sets can share genes, the test results should satisfy certain logical constraints. We will develop computationally efficient statistical methods for multiple testing of these logically constrained hypotheses. As sequencing technology emerges as the primary tool for investigating genome-wide regulation events, it becomes increasingly important to develop statistical methodology for error rate estimation and control for genomic event detection. Finally, as more and more high-throughput genomic data accumulate, we will develop methods for joint analysis of multiple datasets to improve the sensitivity and specificity of genomic discoveries.