Unified Statistical Methods for Sequence-Based Association Studies

描述（由申请人提供）：基于序列的关联研究的统一统计方法。快速和经济的下一代测序（NGS）技术将产生前所未有的大量（数千个个体）和高维（千万）基因组和表观基因组变异数据，允许几乎完整的基因组和表观基因组变异评估，包括常见和罕见变异，RNA-seq， mRNA-seq和甲基化-seq数据。因此，这些基因组变异数据如此密集地分布在整个基因组中，以至于这些遗传变异可以被认为是连续体上的基因组变异观察。NGS技术的出现不仅改变了我们对基因组学的看法，从独立分离的离散模型到混合（离散和连续）模型，而且使基因组和表观基因组分析的分析方法从标准多变量数据分析到功能数据分析，从独立抽样到依赖抽样，从低维数据分析到高维数据分析，都发生了巨大的变化。从单一基因组或表观基因组变异分析到综合基因组和表观基因组分析。为了解决我们在NGS数据分析中面临的巨大挑战，本提案的目标是为基于序列的关联研究和QTL （eQTL）分析开发新颖而强大的统计方法，利用高维数据降维、因果推理和功能数据分析技术来识别基因组中常见和罕见的风险变异，通过中间表型和表达来研究它们的功能。估计变异对表型的总影响（干预影响）和直接影响，并统一基于家族和人群的设计。我们将通过模拟和真实数据集来评估这些方法的性能。

DESCRIPTION (provided by applicant): Unified Statistical Methods for Sequence-based Association Studies. Fast and economic next generation sequencing (NGS) technologies will generate unprecedentedly massive (thousands of individuals) and high-dimensional (ten millions) genomic and epigenomic variation data that allow nearly complete evaluation of genomic and epigenomic variation including common and rare variants, RNA-seq, mRNA-seq and methylation-seq data. As a consequence, these genomic variation data are so densely distributed across the genome that the genetic variants can be considered as genomic variation observations varying over a continuum. The emergence of NGS technologies is not only changing our view of genomics from independently segregating discrete model to hybrid (both discrete and continuous) models, but also causing great changing in analytic methods for genomic and epigenomic analysis from standard multivariate data analysis to functional data analysis, from independent sampling to dependent sampling, from low dimensional data analysis to high dimensional data analysis, from single genomic or epigenomic variant analysis to integrated genomic and epigenomic analysis. To address the great challenges we are facing in NGS data analysis, the goals of this proposal are to develop novel and powerful statistical methods for sequence-based association studies and QTL (eQTL) analysis which leverage high dimensional data reduction, causal inference and functional data analysis techniques to identify both common and rare risk variants across the genome, investigate their function via intermediate phenotypes and expressions, estimate the total effects (intervention effects) and direct effects of variants on the phenotypes, and unify family and population-based designs. We will evaluate the performance of these methods by simulated and real datasets.