Ulysses: accurate detection of low-frequency structural variations in large insert-size sequencing libraries
Ulysses: accurate detection of low-frequency structural variations in large insert-size sequencing libraries
复制标题
DOI：
10.1093/bioinformatics/btu730
复制
发表时间：
2015-03-15
期刊：
影响因子：
5.8
通讯作者：
Lafontaine, Ingrid
中科院分区：
文献类型：
--
作者：
Gillet-Markowska, Alexandre;Richard, Hugues;Lafontaine, Ingrid
动机：在短距离配对端（PE）文库中检测结构变异（SVs）仍然具有挑战性，因为SV断点可能涉及大量分散的重复序列，或者携带固有的复杂性，很难用经典的PE测序数据解决。相比之下，大型插入大小的测序文库（Mate-Pair文库）提供了更高的基因组物理覆盖率，并提供了访问含有重复序列的区域。因此，从理论上讲，它们可以克服以前的限制，因为它们变得日常可用。然而，这种类型的文库通常具有较大的插入大小分布和较高的嵌合序列率，这给SV的准确注释带来了挑战。结果：在这里，我们展示了Ulysses，一个工具，实现了比现有工具更高的检测精度，无论是模拟的还是真实的配对测序数据集，来自1000人类基因组计划。Ulysses通过评估每个可能的变异（重复、缺失、易位、插入和反转）的统计显著性，以原则性的方式实现了对整个变异谱的高特异性，而不是针对产生实验噪声的明确模型。这个统计模型被证明对低频变异的检测特别有用。对来自乳腺癌样本的大量插入物Mate-Pair文库进行的SV检测显示，肿瘤中存在高水平的体细胞复制，在较小程度上，血液样本中也存在。总之，这些结果表明，Ulysses是表征人体组织和癌症基因组中体细胞嵌合体的有价值的工具。
Motivation: The detection of structural variations (SVs) in short-range Paired-End (PE) libraries remains challenging because SV breakpoints can involve large dispersed repeated sequences, or carry inherent complexity, hardly resolvable with classical PE sequencing data. In contrast, large insert-size sequencing libraries (Mate-Pair libraries) provide higher physical coverage of the genome and give access to repeat-containing regions. They can thus theoretically overcome previous limitations as they are becoming routinely accessible. Nevertheless, broad insert size distributions and high rates of chimerical sequences are usually associated to this type of libraries, which makes the accurate annotation of SV challenging.Results: Here, we present Ulysses, a tool that achieves drastically higher detection accuracy than existing tools, both on simulated and real mate-pair sequencing datasets from the 1000 Human Genome project. Ulysses achieves high specificity over the complete spectrum of variants by assessing, in a principled manner, the statistical significance of each possible variant (duplications, deletions, translocations, insertions and inversions) against an explicit model for the generation of experimental noise. This statistical model proves particularly useful for the detection of low frequency variants. SV detection performed on a large insert Mate-Pair library from a breast cancer sample revealed a high level of somatic duplications in the tumor and, to a lesser extent, in the blood sample as well. Altogether, these results show that Ulysses is a valuable tool for the characterization of somatic mosaicism in human tissues and in cancer genomes.