Robust Methods for the Efficient Analysis and Integration of DNA Sequence Data

描述(申请人提供)：人类遗传学研究正处于捕捉遗传变异方式的重大转变的尖端--从基于标记的方法转变为基于通过测序对个人基因组进行完整表征的方法。这是一个令人兴奋的前景，但也不是没有挑战。即将产生的大量序列数据引发了如何最好地使用这些数据的几个问题。例如，由于数据的巨大规模，将序列变异与人类疾病联系起来的统计方法需要在统计和计算上都是有效的。此外，近期内的大多数遗传关联实验将不仅仅依赖于序列数据，而是将拥有具有序列数据的个体的子样本，而样本的其余部分将保持未测序，但将包含基因信息。或者，序列数据可以在单独的外部样本上获得。因此，重要的是制定统计方法，将这些不同类型的数据适当地整合到一个统一的推理框架中。这项研究项目将通过提出开发一类新的基于序列的单倍型共享统计来解决这些问题，该统计在变异/疾病关联测试中利用DNA序列进化的影响(具体目标1)。此外，我们建议开发一个统计框架，允许对DNA序列和基因数据进行统一分析(具体目标2)。在整个过程中，我们将利用我们之前的工作，开发用于单倍型推断的稳健方法，以开发在计算和统计上有效的程序，这些程序对群体遗传假设保持稳健。将强调分层分析方法，以便对人口分层造成的混乱情况进行调整。将提出有效的蒙特卡罗程序来解释所研究的大量序列变体。我们将开发一套软件工具，充分实施所开发的方法，并向一般研究界免费提供这些工具(具体目标3)。最后，使用这些工具，我们将分析一个公开可用的DNA序列数据集，目标是更好地定位与疾病相关的序列变异(特定目标4)。通过这一提议开发的方法代表了一个统一的、统计上严格的框架，用于开发强大的测试，利用DNA序列之间的进化关系，同时允许将不同的数据类型合并到统一的分析中。这些程序将为研究人员提供更精细地定位与疾病相关的序列变异的工具，使变异能够更好地优先于通过功能研究进行后续研究。人类遗传学研究正处于捕捉遗传变异方式的重大转变的尖端--从基于标记的方法转变为基于通过测序对个人基因组的完整表征的方法。然而，即将产生的大量测序数据导致了关于它们的统计分析和纳入更大实验的问题。我们通过提出一个统一的和统计上严格的框架来解决这些问题，以开发强大的测试，该测试利用DNA序列之间的进化关系，并允许将不同的数据类型合并到统一的分析中。

DESCRIPTION (provided by applicant): Human genetics research is on the cusp of a major transformation in how genetic variation is captured-from a marker-based approach to one based on a complete characterization of an individual's genome by sequencing. This is an exciting prospect but not without its challenges. The imminent production of large amounts of sequence data raises several issues on how best to use these data. For example, because of the sheer scale of the data, statistical approaches for associating sequence variants with human disease need to be efficient, both statistically and computationally. In addition, most genetic association experiments in the near term will not rely solely on sequence data but instead will have sub-samples of individuals with sequence data while the rest of the sample will remain unsequenced but will contain genotype information. Alternatively, sequence data may be available on a separate, external sample. Thus it will be important to develop statistical methods that can appropriately integrate these various types of data into a unified inferential framework. This research project will address these issues by proposing to develop a novel class of sequence- based haplotype sharing statistics that exploit the implications of DNA sequence evolution in testing for variant/disease association (specific aim 1). Further, we propose to develop a statistical framework that allows for the unified analysis of DNA sequence and genotype data (specific aim 2). Throughout we will leverage our previous work developing robust methods for haplotype inference to develop computationally and statistically efficient procedures that remain robust to population genetic assumptions. A stratified analytic approach will be emphasized to allow for adjustment for confounding due to population stratification. Efficient Monte Carlo procedures will be proposed to account for the large number of sequence variants investigated. We will develop a suite of software tools that fully implement the methodology developed and make them freely available to the general research community (specific aim 3). Finally, using these tools, we will analyze a publicly available DNA sequence dataset with the goal of better localizing disease- associated sequence variants (specific aim 4). The methods developed through this proposal represent a unified and statistically rigorous framework for developing powerful tests that exploit evolutionary relationships between DNA sequences while allowing for disparate data types to be incorporated into a unified analysis. These procedures will give researchers the tools to more finely localize disease-associated sequence variants, allowing variants to be better prioritized for subsequent investigation via functional studies. Human genetics research is on the cusp of a major transformation in how genetic variation is captured-from a marker based approach to one based on a complete characterization of an individual's genome by sequencing. The imminent production of large amounts of sequencing data, however, leads to questions concerning their statistical analysis and incorporation into the larger experiment. We address these questions by proposing a unified and statistically rigorous framework for developing powerful tests that exploit evolutionary relationships between DNA sequences and that allow for disparate data types to be incorporated into a unified analysis.