Statistical and computational analysis in whole genome sequencing studies.

描述(由申请人提供)：这个项目将调查基于全基因组测序(WGS)的基因组研究的统计和计算分析所产生的几个问题。在WGS研究的数据管理领域，我们解决了与序列读取的海量文件及其相关质量分数的传输和存储相关的快速增长的成本。我们将开发数据压缩方法，以实现比当前标准更高的数倍压缩，并将产生的错误降至最低。在二次分析方面，我们将开发新的统计学习方法，以改进不同质量分数的重新校准，并过滤不可靠的呼叫。这将提高WGS数据提供的关键信息的可靠性，WGS数据是指示基因组与参考不同的位置和差异的性质的变体调用。我们将研究基于WGS的病例对照研究方法。特别是，我们将开发统计模型，以便能够整合来自多种类型变体的信息 以获得更强大的关联性测试。我们将把在这一目标中开发的方法应用于腹主动脉瘤研究中的WGS数据分析。最后，我们将讨论与人口规模WGS项目相关的选定新问题。最近启动了几个国家计划，为数十万有纵向医疗记录的个人生成WGS数据。在人口规模上获得这种全面的数据将为基因组医学开辟一个丰富的前沿，并将为统计分析带来许多新的挑战。我们将制定其中一些新的挑战，并制定应对这些挑战所需的统计方法。

DESCRIPTION (provided by applicant): This project will investigate several issues arising from the statistical and computational analysis of whole genome sequencing (WGS) based genomics studies. In the area of data management in WGS studies, we address the rapidly increasing cost associated with the transfer and storage of the massive files for the sequence reads and their associated quality scores. We will develop data compression methods to achieve a further compression of several folds beyond current standards, with minimal incurred errors. In the area of secondary analysis, we will develop new statistical learning methods to improve variant quality score recalibration and to filter out unreliable calls. This will improve te reliability of the key information provided by the WGS data, which are the variants calls indicating the locations where the genome differs from the reference and the nature of the differences. We will study methods for case-control studies based on WGS. In particular, we will develop statistical models to enable the integrating of information from multiple types of variants to obtain more powerful tests of association. We will apply the methods developed in this aim to the analysis of WGS data from a study on abdominal aortic aneurysm. Finally, we will address selected new questions associated with population scale WGS projects. Several national programs have recently been initiated to generate WGS data for hundreds of thousands of individuals with longitudinal medical records. The availability of this comprehensive data on a population scale will open up a rich frontier for genome medicine and will pose many new challenges for statistical analysis. We will formulate some of these new challenges and develop the statistical methods needed to meet these challenges.