Statistical methods for population and family-based whole-genome sequence data

描述（由申请人提供）： 新兴的测序技术使全基因组测序成为研究各种表型/感兴趣的疾病，特别是关注罕见变异位点的研究。虽然第一批测序项目主要集中在无关个体的分析上，但随着测序成本的迅速降低，最近已经进行或启动了许多包括相关个体的测序研究。然而，分析基于家族的序列数据的方法在很大程度上落后，部分原因是家族结构的复杂性和计算障碍。在这项研究中，我们的主要目标是有效和准确地推断个人基因型和单体型-任何测序项目的关键组成部分-通过结合来自家庭和人口水平的信息，并研究差异测序错误将如何影响下游关联分析。为了实现这些目标，我们提出了具体的目标如下：1）我们将提出一个新的统计框架的基因分型调用和单倍型推断的序列数据，包括相关的个人。新方法利用了两个非相关个体之间共享的短链和家族成员之间共享的长链，同时通过两种经典方法的协同作用保持了高度的准确性：用于连锁不平衡信息的隐马尔可夫模型（HMM）和用于遗传向量的Lander-Green算法; 2）我们将开发一种精确的HMM计算算法，以加速一类广泛使用的遗传程序，包括在Aim 1中开发的方法，而不牺牲精度; 3）我们将评估测序错误对家庭的影响-的关联方法，并使用目标1中提出的方法的内在随机性，以减少假阳性的框架下，多重插补; 4）我们将与正在进行的测序项目合作，测试和重新校准我们开发的方法，并系统地研究不同的研究设计。这些目标的成功完成将产生最先进的统计方法和软件，这将促进快速增长的测序项目，包括家庭成员，并指导未来研究的设计和分析。

DESCRIPTION (provided by applicant): Emerging sequencing technologies have made whole-genome sequencing become available for researches to study various phenotypes/diseases of interest, particularly focusing on rare variants sites. Although the first batch of sequencing projects has mainly focused on the analysis of unrelated individuals, numerous sequencing studies including related individuals have been carried out or launched recently as the sequencing cost reduces rapidly. However, the methodologies for analyzing family-based sequence data are largely falling behind partially due to the complexity of family structures and computational barrier. In this study, our primary goals are to efficiently and accurately infer individual genotypes and haplotypes - the key component of any sequencing project - by combining information from both family and population levels, and to study how differential sequencing errors will affect downstream association analysis. To achieve these goals, we propose specific aims as follows: 1) We will propose a novel statistical framework for genotyping calling and haplotype inference of sequence data including relative individuals. The new method takes advantages of both short stretches shared between unrelated individuals and long stretches shared between family members in a computationally feasible manner while retaining a high degree of accuracy via the synergy between two classic approaches: hidden Markov model (HMM) for linkage disequilibrium information and Lander-Green algorithm for inheritance vectors; 2) We will develop an exact algorithm for HMM computation to speed up a class of widely use genetics programs, including the method developed in Aim 1, without any sacrifice of accuracy; 3) We will assess the impact of sequencing errors on family-based association methods for rare variants and use the intrinsic stochastic nature of the proposed methods in Aim 1 to reduce the false positives under a framework of multiple imputation; 4) We will test and recalibrate our developed methods in collaboration with ongoing sequencing projects and systematically investigate different study designs. Successful completion of these aims will yield state-of-the-art statistical methods and software, which will facilitate the fast growing sequencing projects including family members and guide the design and analysis of future studies.