Methods to enable robust and efficient use of genetic summary data

摘要 公开可用的遗传概要数据对于深入了解遗传病因学具有很高的实用性 健康和疾病。基因频率数据库，如基因组聚合数据库 (GnomAD)，用于区分推定的因果变异的优先顺序，最近，作为 病例对照分析。全基因组关联研究(GWAS)测试统计数据用于各种 二次数据分析包括多基因风险评分(PR)、遗传相关分析和FINE 因果变异的图谱。与个体水平的数据相比，遗传汇总数据往往较少 访问障碍，促进这些宝贵数据资源的广泛使用。提供和使用 摘要遗传数据在所有祖先群体中往往是不公平的，特别是对于研究不足的群体。 在这些资源中几乎没有代表性的祖先群体。此外， 摘要数据中的异质性可能会导致混淆和降低病例对照的能力 分析，对罕见疾病的推定因果变量的优先顺序不正确，以及 多基因风险分数。我开发了健壮而高效的方法来适当地使用遗传总结 数据，同时估计、建模和利用内部的异质性。我的方法结合在一起 一个统一的框架，在这里我翻转了处理遗传变异的遗传和基因组数据的范例 或元素作为我们分析数据而不是个体的观察单位。这 简单而创新的范式转换使经典统计技术的使用和创新成为可能 检测、调整、甚至使用汇总级数据中的异构性的方法。要启用 广泛和公平地使用我们的方法，我们将创建与以下内容兼容的公开可用的R包 用于交互式互联网使用的BioConductor和闪亮应用程序。

Abstract Publiclyavailable genetic summary data canhave high utility for providing insight into genetic etiology of health and disease. Databases of genotype frequencies, such as the genome Aggregation Database (gnomAD), are used to prioritize putative causal variants and, more recently, as pseudo-controls in case-control analysis. Genome Wide Association Study (GWAS) test statistics are used in a variety of secondary data analyses including polygenic risk scores (PRS), genetic correlation analysis, and fine mapping of causal variants. Compared with individual level data, genetic summary data often has fewer barriers in access, promoting broad use of these valuable data resources. The availability and use of summary genetic data is often not equitable across all ancestral groups, especially for understudied ancestral groups that have little to no representation within these resources. Furthermore, heterogeneity within the summary data can lead to confounding and reduced power for case-control analysis, incorrect prioritization of putative causal variants for rare diseases, and reduced accuracy for polygenic risk scores. I develop robust and efficient methods to appropriately use genetic summary data while estimating, modeling, and harnessing the heterogeneity within. My methods coalesce around a unifying framework where I flip the paradigm of genetic and genomic data treating the genetic variant or element as the observational unit by which we analyze the data rather than the individual. This simple, yet innovative paradigm shift enables the use of classical statistical techniques and the creation of methods that detect, adjust for, and even use heterogeneity within summary level data. To enable broad and equitable use of our methods, we will create publicly available R packages compatible with Bioconductor and Shiny Apps for interactive internet use.