De novo inference of stratification and local admixture in sequencing studies.
De novo inference of stratification and local admixture in sequencing studies.
复制标题
DOI：
10.1186/1471-2105-14-s5-s17
复制
发表时间：
2013
影响因子：
3
通讯作者：
Zhang Y
中科院分区：
文献类型：
--
作者：
Zhang Y
人口结构和基因组当地血统的分析在人口和疾病遗传学中变得越来越重要。随着下一代测序技术的进步，个体基因组中完整的遗传变异被快速生成，为了解群体进化历史和在SNP分辨率下识别局部遗传特征提供了前所未有的机会。这些研究的成功很大程度上依赖于能够充分利用测序信息的准确而强大的计算工具。尽管已经开发了许多用于群体结构推断和混合映射的算法，但其中许多仅适用于基因型或单倍型格式的独立 SNP，并且需要大量参考个体。在本文中，我们提出了一种新的概率方法来检测种群结构和局部混合。该方法输入测序数据、基因型数据和单倍型数据。该方法通过单倍型分割来表征遗传变异的依赖性，使得测序研究中检测到的所有变异都可以充分用于推理。该方法进一步利用无限状态贝叶斯马尔可夫模型来执行从头分层和混合推理。使用来自 HapMapII 和 1000Genomes 的模拟数据集，我们表明我们的方法比几种现有算法表现更好，特别是当参考个体有限或没有可用时。我们的方法不仅适用于人类研究，还适用于其他感兴趣的物种的研究，而这些物种的参考信息很少。软件可用性：http://stat.psu.edu/~yuzhang/software/dbm.tar
Analysis of population structures and genome local ancestry has become increasingly important in population and disease genetics. With the advance of next generation sequencing technologies, complete genetic variants in individuals' genomes are quickly generated, providing unprecedented opportunities for learning population evolution histories and identifying local genetic signatures at the SNP resolution. The successes of those studies critically rely on accurate and powerful computational tools that can fully utilize the sequencing information. Although many algorithms have been developed for population structure inference and admixture mapping, many of them only work for independent SNPs in genotype or haplotype format, and require a large panel of reference individuals. In this paper, we propose a novel probabilistic method for detecting population structure and local admixture. The method takes input of sequencing data, genotype data and haplotype data. The method characterizes the dependence of genetic variants via haplotype segmentation, such that all variants detected in a sequencing study can be fully utilized for inference. The method further utilizes a infinite-state Bayesian Markov model to perform de novo stratification and admixture inference. Using simulated datasets from HapMapII and 1000Genomes, we show that our method performs superior than several existing algorithms, particularly when limited or no reference individuals are available. Our method is applicable to not only human studies but also studies of other species of interests, for which little reference information is available. Software Availability: http://stat.psu.edu/~yuzhang/software/dbm.tar