Scalable methods for identity by descent

摘要 在过去的几年里，大型的基因分型队列正在变得更大。我们正在接近一个基因分型的时代 有很大一部分人口的信息可用。信息学的方法是非常需要的 将这些新信息转化为对人类遗传学的洞察。在信息学创新的推动下， 在过去的3年里，IBD节段检测的格局发生了变化。2019年，我们发布了RAPID， 第一个IBD片段调用方法对于生物库规模的队列足够有效。之后，一代人 已经开发了方法来为IBD段调用提供解决方案。此外，我们还提供了新的算法 以及丰富了PBWT数据结构的方法。此外，大规模召回IBD细分市场的影响 通过在一般人群队列中对多样性的精确表征的能力来证明队列， 种群历史和人类行为的研究，基于IBD的亲缘关系估计，以及IBD图谱。 然而，目前从生物库规模的队列中识别IBD节段的成功只是个开始。更多 为了充分释放基因信息的力量，需要信息学方法的发展。第一，当前 方法主要针对较长的IBD节段(大于3或5厘米脏器(Cm))，以及检测能力 对于较短的细分市场来说，这是不够的。此外，所有基因组区域和所有区域的准确率并不是一致高的 人口。其次，目前的方法主要是针对在一对单倍型之间共享的IBD片段。使用 大样本量、多途径的IBD无处不在，但研究不足。第三，IBD的识别方法 需要查询单倍型和参照板(1-vs-n)之间的片段。对于小样本，甚至 针对面板的1-vs-n查询将支持强大的解释，利用 参考面板。然而，目前的IBD分段检测方法主要是批量调用模式， 进行n-vs-n比较，因此不够灵活，无法满足此类需求。在这场竞争性的更新中 项目，我们建议进一步开发高效、准确和灵活的IBD节段检测算法 生物库规模的大型数据。我们将改进跨基因组、跨长度谱的IBD分段呼叫，以及 跨种族；我们将开发多路IBD聚集性检测方法；我们将开发参考- 基于IBD的调用和线程化方法。这些新的信息学方法将使社区能够更好地 利用大型基因分型队列中的遗传关系进行基因发现。

Abstract In the past few years, large genotyped cohorts are getting larger. We are getting closer to the era where genotype information of a large portion of the population is available. Informatics methods are critically needed for translating the new information into insights for human genetics. Powered by informatics innovations, the landscape of IBD segment detection has been transformed in the past 3 years. In 2019, we published RaPID, the first IBD segment calling method efficient enough for biobank-scale cohorts. Afterwards, a generation of methods has been developed to offer solutions for IBD segment calling. In addition, we delivered new algorithms and methods that enriched the PBWT data structure. Also, the impacts of calling out IBD segments in large cohorts are demonstrated by the powering of precision characterization of diversity in general population cohorts, the studies of population history and human behavior, IBD-based relatedness estimate, and IBD-mapping. However, current success in identifying IBD segments from biobank-scale cohorts is only the beginning. More informatics method developments are needed to fully unleash the power of genotype information. First, current methods are mainly for longer IBD segments (greater than 3 or 5 centimorgans (cM)), and the detection power for shorter segments are insufficient. Also the accuracy is not uniformly high across all genomic regions and all populations. Second, current methods are mainly for IBD segments shared between a pair of haplotypes. With large sample sizes, multi-way IBDs are omnipresent but under-studied. Third, methods for identifying IBD segments between a query haplotype and reference panels (1-vs-n) are needed. For a small sample or even individuals, 1-vs-n query against a panel will enable powerful interpretation leveraging the rich information in the reference panel. However, current IBD segment detection methods are mainly a batch calling mode that conducts n-vs-n comparisons and thus are not flexible enough to address such needs. In this competitive renewal project, we propose to further develop efficient, accurate, and flexible algorithms for IBD segment detection for large biobank-scale data. We will improve IBD segment calling across the genome, across length-spectrum, and across ethnicities; we will develop methods for multi-way IBD cluster detection; and we will develop reference- based IBD calling and threading methods. These new informatics methods will enable the community to better leverage the genetic relationships in large genotyped cohorts for genetic discovery.