Advanced strategies for genotype imputation

描述(由申请人提供)：最近的全基因组关联(GWA)研究发现了许多与疾病易感性有关的等位基因。基因分型方法是这一成功的关键因素。这些统计方法利用公开可用的参考面板中的密集基因类型来估计GWA研究中数百万个未测量的遗传标记的基因类型。因此，它们使研究人员能够测试更多的疾病关联标记物，而不是那些已被实验测量的标记物，从而提高检测风险变量的能力。随着新一代测序技术的出现，将有助于检测罕见的基因变异与疾病的关联，归因于疾病的重要性可能只会增加。然而，在优化补偿方法的应用方面仍有一些挑战尚未解决。虽然推算的准确性取决于使用适当的参考个人，但关于如何以最佳方式选择用作模板的个人的数据有限，特别是在非洲裔美国人和西班牙裔/拉丁裔混合人口中。此外，主要针对常见的遗传变异评估了补偿算法的性能。随着遗传学研究开始关注罕见变异作为不明原因遗传病风险的潜在重要来源，改善这种多态的基因归因的性质是至关重要的。为解决这些问题提出了四个项目。首先，将使用多个现有的参考数据集、推算算法和推算准确度测量来评估非洲裔美国人和西班牙裔/拉丁裔人口的推算准确性和统计能力。该项目将通过优化非洲裔美国人和西班牙裔/拉丁裔人口的归因，促进确定这些人口中的疾病易感基因。其次，通过考虑混杂个体基因组的独特马赛克结构，将设计出新的基于模型的统计技术来进行归罪。这项工作建立在流行的FAST PHASE软件的基础上，以进一步加强混合种群的归因。第三，将设计和测试稀有变异的方法，包括拷贝数变异。这项分析将使在GWA测试中使用罕见的变种成为可能，从而改善揭示其对疾病风险的影响的前景。第四，将开发算法来最佳地选择用于重测序的个体，并将其用作模板个体以进行归罪。这项工作将加强即将进行的GWA研究的设计，该研究将纳入样本子集的重新测序数据。这些项目将通过模拟、理论和计算分析相结合的方式完成。此外，算法将使用来自巴尔的摩的非裔美国人、来自德克萨斯州斯塔尔县的墨西哥裔美国人和1000基因组计划的数据集来应用。该项目产生的统计资源将通过公开可用的软件传播，将提供必要的工具，以促进目前正在进行的绘制疾病基因图谱的工作，特别是在非洲裔美国人和西班牙裔/拉美裔人口中。 与公共健康相关：许多疾病基因已经被“关联研究”确定，这些研究在人类基因组中搜索基因变异，这些变异在疾病携带者中比对照个体中出现得更频繁。我们将通过确定将遗传相关性研究的数据与现有数据库的数据相结合的最佳统计策略，来改善识别疾病基因的前景。我们的项目将提供关于最佳研究特征和统计方法的指导方针，以在非洲裔美国人和墨西哥裔美国人等研究不足、信息丰富的人群中寻找疾病基因。

DESCRIPTION (provided by applicant): Recent genome-wide association (GWA) studies have identified many alleles contributing to disease susceptibility. Genotype imputation methods have been a key contributor to this success. These statistical approaches leverage dense genotypes in publicly available reference panels to estimate genotypes at millions of unmeasured genetic markers in a GWA study. Thus, they enable investigators to test many more markers for disease association beyond those that have been experimentally measured, thereby improving power to detect risk variants. With the recent advent of next-generation sequencing technologies that will facilitate the testing of rare genetic variants for disease association, the importance of imputation is only likely to increase. However, several challenges for optimizing the application of imputation methods remain unaddressed. While imputation accuracy depends on the use of appropriate reference individuals, limited data exist on how to optimally choose the individuals used as a template, particularly in admixed populations such as African Americans and Hispanic/Latino populations. Moreover, the performance of imputation algorithms has been evaluated primarily for common genetic variants. As genetic studies begin to focus on rare variation as a potentially important source for unexplained heritable disease risk, it is essential to improve the properties of genotype imputation for such polymorphisms. Four projects are proposed for addressing these issues. First, imputation accuracy and statistical power will be evaluated in African Americans and in a Hispanic/Latino population, using multiple existing reference datasets, imputation algorithms, and imputation accuracy measures. This project will facilitate the identification of disease-susceptibility loci in African Americans and Hispanic/Latino populations by optimizing imputation in these populations. Second, new model-based statistical techniques for imputation will be devised by considering the unique mosaic structure of genomes of admixed individuals. This work builds on the popular fastPHASE software to further enhance imputation in admixed populations. Third, methods of imputing rare variants, including copy-number variants, will be devised and tested. This analysis will enable the use of rare variants in GWA tests, thereby improving the prospects for uncovering their effects on disease risk. Fourth, algorithms will be developed for optimally selecting individuals for resequencing and use as template individuals for imputation. This work will enhance the design of forthcoming GWA studies that will incorporate resequencing data on subsets of the sample. The projects will be accomplished through a combination of simulation, theory, and computational analysis. Furthermore, algorithms will be applied using datasets on African Americans from Baltimore, Mexican Americans from Starr County, Texas, and the 1000 Genomes Project. Statistical resources generated from the project, which will be disseminated in publicly available software, will provide essential tools for facilitating the ongoing effort of mapping disease genes, particularly in African Americans and Hispanic/Latino populations. PUBLIC HEALTH RELEVANCE: Many disease genes have been identified by "association studies" that search the human genome for genetic variants that occur more frequently in individuals who carry a disease than in control individuals. We will improve the prospects for identifying disease genes by determining the best statistical strategies for combining data from genetic association studies with data from existing databases. Our project will provide guidelines about optimal study characteristics and statistical methods to find disease genes in understudied, informative populations such as African Americans and Mexican Americans.