Algorithmic strategies for detecting structural variation in genomes

描述（由申请人提供）：精细尺度的核苷酸变化以及基因重组通常被认为是人类遗传变异的主要来源[1,13,14]。对更大规模（bbb10kb）的基因组结构变异知之甚少。随着基因组技术的进步，我们正在检测越来越多的结构变异，包括基因组倒置[24,48,71,65,31]；插入/删除多态性[12,26,42]；拷贝数多态性[28,59,60]。这些大的变异可以完全破坏基因的编码和调控位点以及拷贝数，从而对人类表型和疾病易感性产生巨大影响[23,61]。在癌症和其他疾病中确实观察到有害影响[70,43]。我们对这些变化的规模和影响的理解可以通过改进从这些技术中挖掘数据的计算工具来增强。在这里，我建议发展算法和计算工具，以提高结构变化的检测和分辨率（断点的位置）。具体来说，我将开发算法(a)测序项目的实验设计，用于检测和解决结构变化；(b)使用末端序列分析对断点进行精细定位，以检测基因破坏和基因融合；(c)重建肿瘤基因组结构；(d)通过多重PCR检测正常细胞与突变细胞异质混合中的靶向基因组变异；(e)基因型数据中平衡结构变异的检测。这些工具将使用统计机器学习和组合算法的技术来设计。验证将使用已知的结构变化、模拟研究以及与技术开发人员和早期技术采用者的广泛实验合作进行。所有的数据和软件将免费提供给学术和非商业用途。

DESCRIPTION (provided by applicant): Fine-scale nucleotide changes, along with genetic recombination, are often cited as the major source of human genetic variation [1, 13, 14]. Less is known about larger scale (> 10kb) genomic structural variations. As genomic technologies improve, we are detecting structural variation in ever-increasing numbers, including genomic inversions [24, 48, 71, 65, 31]; insertion/deletion polymorphisms [12, 26, 42]; and, copy number polymorphisms [28, 59, 60]. These large variations can completely disrupt coding and regulatory sites and copy number of genes, and thereby have a huge impact on human phenotypes and disease susceptibility [23, 61]. Deleterious effects have indeed been observed in cancer and other diseases [70, 43]. Our understanding of the scale and impact of these variations can be enhanced by improving computational tools for mining the data from these technologies. Here, I propose the development of algorithms and computational tools to improve detection and resolution (location of breakpoints) of structural variation. Specifically, I will develop algorithms for (a) experimental design of sequencing projects for detecting and resolving structural variations; (b) fine-mapping of breakpoints using end sequence profiling, to detect gene-disruption and gene-fusions; (c) reconstructing tumor genome architectures; (d) detection of targeted genomic variations in a heterogeneous mix of normal versus mutated cells via multiplex PCR; and (e) detection of balanced structural variation in genotype data. The tools will be designed using techniques from statistical machine learning and combinatorial algorithms. Validation will be performed using known structural variations, simulation studies, and extensive experimental collaborations with technology developers and early technology adopters. All of the data, and software will be freely available for academic and non-commercial uses. PUBLIC HEALTH RELEVANCE: The proposed computational tools will be used to detect structural variations in human populations as a starting point for understanding their role in normal evolution and disease, specifically cancer. The architecture of tumor genomes will help reveal genes that are disrupted and differentially expressed in tumor cells. The targeted detection of genomic lesions in a heterogeneous mix of mutated and wildtype cells, will find application as an early diagnostic for cancer. Thus, our computational methods will have an immediate and long term effect on human health.