Increasingly efficient methods for inferring the ancestral origin of genome regions are needed to gain new insights into genetic function and history as biobanks grow in scale. Here we describe two near-linear time algorithms to learn ancestry harnessing the strengths of a Positional Burrows-Wheeler Transform (PBWT). SparsePainter is a faster, sparse replacement of previous model-based `chromosome painting' algorithms to identify recently shared haplotypes, whilst PBWTpaint uses further approximations to obtain lightning-fast estimation optimized for genome-wide relatedness estimation. The computational efficiency gains of these tools for fine-scale local ancestry inference offer the possibility to analyse large-scale genomic datasets in completely novel ways. Application to the UK Biobank shows that haplotypes better represent ancestries than principal components, whilst linkage-disequilibrium of ancestry identifies signals of recent changes to population-specific selection for many genomic regions associated with immune responses, suggesting new avenues for understanding the pathogen-immune system interplay on a historical timescale.

中文翻译：

---

基于稀疏单倍型的大规模局部血统推断揭示了最近对免疫反应的选择

随着生物样本库规模的扩大，需要越来越有效的方法来推断基因组区域的祖先起源，以获得对遗传功能和历史的新见解。在这里，我们描述了两种利用位置 Burrows-Wheeler 变换 (PBWT) 的优势来学习祖先的近线性时间算法。SparsePainter 是以前基于模型的“染色体绘制”算法的更快、稀疏的替代品，用于识别最近共享的单倍型，而 PBWTpaint 使用进一步的近似来获得针对全基因组相关性估计而优化的闪电快速估计。这些用于精细本地祖先推断的工具的计算效率增益提供了以全新方式分析大规模基因组数据集的可能性。英国生物库的应用表明，单倍型比主要成分更好地代表祖先，而祖先的连锁不平衡识别了与免疫反应相关的许多基因组区域的群体特异性选择的近期变化信号，这为理解病原体免疫系统提供了新途径在历史时间尺度上相互作用。