Computational methods for large-scale genotype data

项目摘要 基因数据集的大小正在呈指数级增长。以目前的增长速度，最大的 在5-7年内，阶段性的、已排序的个体的参考小组将拥有数百万个体。这 研究将解决在以下方面执行基因型阶段化和推算的计算挑战 大的队列和大的参考小组。 来自异种繁殖群体的大群体通常包含名义上无关和密切相关的混合体 相关的个人。这些大型数据集的当前分阶段方法不能对父子代或 其他亲密关系。我们将开发一种新的定相方法，大大提高相位精度。 在密切相关的个体中，这可以扩展到大样本量。 增加参照板的大小也会增加基因型阶段和归因的准确性。然而， 计算成本也随着参考板大小的增加而增加。我们将开发一种新的参考文件格式 这极大地减少了大量引用的补偿和阶段化的计算成本 电池板。我们将提供格式规范、软件和软件库，以便其他研究人员 并且软件开发人员可以很容易地使用新的参考文件格式。 我们将开发一种新的计算方法来寻找共享单倍型片段 参考面板和目标单倍型。这种新方法将显著降低分阶段的成本。 并使用大型参照板进行归责。 最后，我们将推广最快、最准确的基因分相和归类方法(Beagle 5.0)对X染色体数据进行分析。这一扩展将改善对这一重要基因的研究 染色体。

Project Summary The size of genetic data sets is growing exponentially. At the current rate of growth, the largest reference panels of phased, sequenced individuals will have millions of individuals within 5-7 years. This research will address the computational challenges of performing genotype phasing and imputation in large cohorts and with large reference panels. Large cohorts from outbred populations typically contain a mixture of nominally unrelated and closely related individuals. Current phasing methods for these large data sets do not model parent-offspring or other close relationships. We will develop a new phasing method that greatly increases phase accuracy in closely-related individuals and that scales to large sample sizes. Increasing reference panel size also increases genotype phase and imputation accuracy. However, computational cost also increases with reference panel size. We will develop a new reference file format that substantially reduces the computational cost of imputation and phasing with large reference panels. We will provide a format specification, software, and software libraries so that other researchers and software developers can readily use the new reference file format. We will develop a new computational method for finding shared haplotype segments between a reference panel and a target haplotype. This new method will significantly reduce the cost of phasing and imputation using large reference panels. Finally, we will extend the fastest, most accurate method for genotype phasing and imputation (Beagle 5.0) to analyse chromosome X data. This extension will improve genetic studies of this important chromosome.