A Novel Trio-based Bayesian Method to Identify Rare Variants for Birth Defects

 描述(由申请人提供)：尽管全基因组和候选基因关联研究的进展继续发现导致出生缺陷的常见遗传变异(婴儿死亡的主要原因[54])，但最近的证据表明，已发现的基因座只占风险的一小部分。因此，研究人员正在将重点转移到研究可能导致出生缺陷缺失遗传性的罕见变异上。三组设计(对受影响的儿童和父母双方进行基因分型)是此类关联研究的首选，因为它对种群亚结构很稳健，只需要对三个人进行测序，而不是更大的家庭。然而，现有的使用TRIO数据分析常见变体的方法在使用序列数据分析稀有变体时受到功率降低的阻碍。因此，迫切需要开发一种方法，利用TRIO设计联合测试常见和罕见的变异，以全面模拟与出生缺陷相关的遗传因素，以便识别驱动与疾病风险关联的遗传变异。目前分析稀有变异的方法主要集中在汇集一个地区的稀有变异，并对该地区进行全球测试。为了进一步提高计算能力，一些方法在模型中加入了常见变量作为协变量。在为数不多的能够从统计上识别导致这种关联的罕见变异的最新方法中，都是基于病例对照设计，该设计要求一直在收集三组数据的出生缺陷研究人员找到一组外部对照。这并不理想。迫切需要一种方法，可以使用Trio数据以统计方式识别导致出生缺陷相关的遗传变异(罕见和常见)，从而联合分析常见和罕见的变异。为了满足这一需求，我们提出了以下目标：(1)开发一种贝叶斯随机搜索变量选择方法，用于使用TRIO数据进行常见和罕见的变量分析，该方法可以识别导致关联的变量；以及(2)使用先天性心脏缺陷的模拟和真实数据，将全局分量的性能与现有的TRIO数据的全局测试进行比较。为了实现这些目标，我们将为三个数据开发一个新的贝叶斯模型，使用期望最大化算法来快速计算贝叶斯后验分布的模式，作为对遗传区域以及特定变体(罕见和常见)的估计。拟议的研究具有创新性，因为它将发展新的和强大的统计方法。这项拟议的研究意义重大，因为它将产生一种新的强大且广泛适用的方法来揭示出生缺陷和其他儿童疾病的遗传基础。最终，新方法将改善出生缺陷的研究，应用于真实数据可以提高我们对先天性心脏病病因的理解，潜在地改善先天性心脏病的遗传咨询和预防策略。

 DESCRIPTION (provided by applicant): Although advances in genome wide and candidate gene association studies continue to identify common genetic variants contributing to birth defects (a leading cause of infant mortality [54]), recent evidence indicates that discovered loci only account for a small fraction of risk. As a result, researchers are shifting focus to investigae rare variants that could be responsible for this missing heritability of birth defects. The trio design (genotyping the affected child and both parents) is preferred for such association studies because it is robust to population substructure and only requires sequencing three people, rather than larger families. However established methods for analyzing common variants using trio data are hindered by reduced power when analyzing rare variants using sequence data. Thus, there is a critical need to develop methods that jointly test for common and rare variants using the trio design to comprehensively model the genetic factors associated with birth defects in order to identify the genetic variants driving the association with disease risk. Current methods for analyzing rare variants primarily focus on pooling the rare variants in a region and performing a global test for that region. To further increase power, some methods include common variants in the model as covariates. Of the few most recent methods that can statistically identify rare variants that drive the association, all are based on the case control design, which requires birth defect researchers who have been collecting trio data to find an external set of controls. This is less than ideal. There is a strong need for methods that can jointly analyze common and rare variants using trio data in such a way that can statistically identify the genetic variants (rare and common) that drive association with birth defects. We propose to fulfill this need with the following aims: (1) Develop a Bayesian stochastic search variable selection method for common and rare variant analysis using trio data that can identify the variants driving the association; and (2) Compare the performance of the global component to existing global tests for trio data using both simulated and real data for congenital heart defects. To achieve these aims, we will develop a novel Bayesian model for trio data, using the Expectation-Maximization algorithm to quickly compute the modes of the Bayesian posterior distribution as estimates for genetic regions as well as specific variants (rare and common). The proposed research is innovative because it will develop new and powerful statistical methodology. The proposed research is significant because it will produce a new powerful and widely applicable method for uncovering the genetic basis for birth defects and other childhood diseases. Ultimately, the new methods will improve birth defects research, and the application to real data can improve our understanding of the etiology of congenital heart defects, potentially improving genetic counseling and prevention strategies for congenital heart defects.