Statistical methods to infer structure and impact of ancient admixture

描述(申请人提供)：过去十年对人类遗传变异的研究表明，高度分化的人口群体(古老的混合体)之间的混合事件，如尼安德特人和非非洲人之间的混合事件一直是常见的，并可能对人类表型产生重大影响。例如，研究已经证明了它在分析单个基因座(如MHC基因座)时的表型影响。然而，缺乏足够的分析工具阻碍了对古代混合物的表型影响的系统了解。这项K99/R00研究建议开发和验证统计方法来推断古混合物产生的遗传结构，并利用这一结构来识别由古混合物引入的影响表型的遗传变异。应用这些方法的洞察力不仅将产生对常见疾病等复杂表型背后的遗传因素的更全面的了解，而且还将确保目前服务不足的少数群体--其中许多人是混合事件的后裔或来自与欧洲人不同的祖先群体--可以像欧洲人后裔一样得到有效的研究，并可以从基因组医学的发现中受益。这项建议的第一个目标是扩展和验证我们目前的统计模型，以准确地推断古老混合物中的本地祖先。提出的模型试图使用条件随机场(CRF)的统计框架来整合大量的遗传变异模式。应用这一模型的第一个重要例子是推断非非洲人口中尼安德特人的当地血统。推断的尼安德特人祖先将被用于第二个目标：将尼安德特人的变体与特定的表型联系起来。这一目标将通过分析旨在捕获尼安德特人衍生变体的定制阵列来实现，并通过扩展CRF来根据SNP基因分型阵列而不是下一代测序来推断尼安德特人的祖先。将探索一种补充的方法来研究自然选择对尼安德特人变异的作用，使用一种新的基于扩散过程的统计检验。最后，CR将被推广到处理多个祖先群体以及没有祖先群体可用的参考基因组的情况，并将在每个案例的重要例子上进行测试和验证，例如将Denisovan混入有未知古老祖先证据的美拉尼西亚群体和撒哈拉以南非洲群体。这项研究的所有方法和结果都将公开。

DESCRIPTION (provided by applicant): Studies of human genetic variation over the last decade have revealed that mixture events between highly diverged population groups (archaic admixture), such as between Neandertals and non-Africans, have been a common occurrence and are likely to have had a major impact on human phenotypes. For example, studies have documented its phenotypic impact in analyses of individual loci, such as the MHC locus. However, a lack of adequate analytical tools has hindered a systematic understanding of the phenotypic impact of archaic admixture. This K99/R00 research proposal proposes to develop and validate statistical methods to infer the genetic structure arising from archaic admixture and to leverage this structure to identify genetic variants introduced by archaic admixture that influence phenotypes. Insights from the application of these methods will not only produce a more complete understanding of the genetic factors underlying complex phenotypes, such as common diseases, but will also ensure that currently under-served minority populations, many of whom descend from admixture events or from ancestral groups distinct from those of Europeans, can be studied just as effectively as populations of European descent and can benefit from the discoveries of genomic medicine. The first goal of this proposal is to extend and validate our current statistical model for accurate inference of local ancestry in archaic admixtures. The proposed model attempts to integrate a large number of patterns of genetic variation using the statistical framework of Conditional Random Fields (CRF). An important first example for the application of this model is the inference of Neandertal local ancestry in non-African populations. The inferred Neandertal ancestry will be leveraged for the second goal: to associate Neandertal variants with specific phenotypes. This goal will be pursued by analyzing a custom array designed to capture Neandertal-derived variants and by extending the CRF to infer Neandertal ancestry from SNP genotyping arrays rather than from next-generation sequencing. A complementary approach to study the action of natural selection on Neandertal variants, using a novel diffusion process-based statistical test, will be explored. Finally, the CR will be generalized to handle multiple ancestral populations as well as to the case where no reference genomes are available for the ancestral populations, and will be tested and validated on important examples for each case such as Denisovan admixture into Melanesian populations and sub-Saharan African populations that have evidence of unknown archaic ancestry. All of the methods and the results from this research will be made publicly available.