Scalable methods for identity by descent

摘要 在接下来的几年里，大的基因分型队列变得可用（例如，TOPMed，英国生物银行，我们所有人， 百万退伍军人计划）。随着样本量接近总人口规模的0.1%-1%， 远亲和血统认同或IBD信息在这样的样品中表现。此类信息 将使更复杂和强大的遗传学分析超越单一的变异为基础的分析。然而，在这方面， 目前的信息学方法不具备处理该规模的基因型数据的效率。我们将 为具有基因型的生物库规模队列开发新的基因组信息学方法。我们已经开发了一个 高效的工具，RaPID，第一个计算上可行的方法推断IBD片段之间的个人， 一个生物库规模的队列我们证明，RaPID实现了运行时间与样本量的线性关系， 比现有方法快100倍以上。同时，RaPID检测到更多数量的IBD， 更高的准确性和更清晰的段边界比现有的方法。在本申请中，我们建议 开发（1）用于成对IBD检测的RaPID+方法，该方法可以容忍和纠正相位误差， 参数调整的原则性方式，并且可以跨测序和阵列平台与基因型数据一起工作; (2)用于检测IBD 2片段的RaPID-二倍体方法;（3）鉴定IBD 2片段的RaPID-多途径方法， IBD集群;（4）RaPID-祖先的方法，在次大陆人群的地方祖先推断。 方法将严格测试模拟使用现实的人口统计模型以及真实的 来自大型队列的数据。所有的方法都将作为学术使用的免费软件来实现。该项目将 通过开发有效的信息学工具来推进遗传研究， 非常大的基因分型队列。

ABSTRACT In the next a few years, large genotyped cohorts are becoming available (e.g., TOPMed, UK biobank, All of Us, Million Veteran Program). With the sample size approaches 0.1%-1% of the total population size, extensive distant relatives and Identity-by-descent, or IBD information are represented in such samples. Such information will enable more sophisticated and powerful genetics analysis beyond single variant-based analyses. However, current informatics methods are not equipped with the efficiency to handle genotype data of that scale. We will develop new genome informatics methods for biobank-scale cohorts with genotypes. We have developed an efficient tool, RaPID, the first computationally feasible method for inferring IBD segments among individuals in a biobank-scale cohort. We demonstrated that RaPID achieves running time linear to the sample size and is over 100 times faster than existing methods. At the same time, RaPID detects a greater number of IBDs, with higher accuracy, and sharper segment boundaries than existing methods. In this application, we propose to develop (1) the RaPID+ method for pairwise IBD detection that can tolerate and correct phasing errors, with a principled way of parameter tuning, and can work with genotype data across sequencing and array platforms; (2) the RaPID-diploid method for detection of IBD2 segments; (3) the RaPID-multiway method that identifies IBD Cluster; and (4) the RaPID-ancestry method for local ancestry inference across subcontinental populations. Methods will be rigorously tested in simulations using realistic population demographic models as well as real data from large cohorts. All methods will be implemented as free software for academic use. This project will advance genetic research by developing efficient informatics tools that reveal detailed genetic relationships in very large genotyped cohorts.