Advanced development of Lancet, an emerging tool for complex variant calling in cancer genomics

摘要 癌症基因组学中的中心挑战之一是准确检测肿瘤细胞中的体细胞突变的能力。 异质性肿瘤，并精确地确定其克隆起源和进化。这些基本知识 是发现新的癌症疗法的关键近年来，全基因组成本的降低， 全外显子组测序使研究人员能够以前所未有的细节解决这些问题。 然而，该领域的一个主要限制是缺乏用于变体调用的方法， 鉴定简单的单核苷酸变体（SNV）和小的插入缺失，以允许表征复杂的 在肿瘤发生和癌症进展中也起重要作用的结构变化。插入缺失大于 用通常使用的基于双折射的方法发现一些碱基是有挑战性的。此外，大多数变体 调用者分别分析肿瘤和正常数据，这可能会引入假阳性，例如当突变发生时， 在正常样本中显示部分支持。为了克服这些缺点，我们最近推出了 Lancet，在ITCR R21的主持下开发的新的体细胞变异识别器。Lancet利用当地 使用区域聚焦彩色de Bruijn图对肿瘤-正常配对数据进行组装和联合分析， the-fly重复序列组成分析和自调整k-mer策略。这导致相对减少的参考 偏倚;检测与参考染色体显著偏离的变异的能力提高 表示;分析规模的减少，导致检测的能力和灵敏度增加 通过本地化、全面的图形探索实现变体;以及动态调整调用行为 根据每个基因组区域的序列情况，在测试中，Lancet显示出上级性能， 在准确性方面，特别是在检测“过渡区”插入缺失方面，所有主要的基于插入缺失的方法（30- 250 bp）。鉴于其在十多个高影响力出版物中的持续采用和成功应用， 柳叶刀准备进行更先进的开发，以使其变异识别能力得到持续改进， 精度和分析能力。具体来说，Lancet目前受到运行时间长于alignment的限制- 基于方法，较长插入的灵敏度降低，缺乏彩色de Bruijn的交互式可视化 图表，以及无法共同分析纵向数据。为了解决这些限制，我们提出以下建议 具体目标：1）提高计算性能，促进用户采用和第三方开发; 2)添加新功能和增强功能，以改进变体检测、定相和数据可视化;以及3） 启用纵向数据的联合装配和分析。影响：通过额外的开发，下一次迭代 将采用先进的算法，快速，高效，准确，本地化，用户友好，应用- 对全基因组时间序列数据进行可修改的变异分析，使其成为领先的方法之一 for variant变异calling调用in cancer癌症research研究.

ABSTRACT One of the central challenges in cancer genomics is the ability to accurately detect somatic mutations in heterogeneous tumors, and precisely determine their clonal origin and evolution. This fundamental knowledge is central to the discovery of new cancer therapies. In recent years, reductions in the cost of whole-genome and whole-exome sequencing have enabled researchers to address these questions in unprecedented detail. However, a major limitation in the field has been a paucity of methods for variant calling that extend beyond identifying simple single-nucleotide variants (SNVs) and small indels to allow the characterization of complex structural changes that also play a significant role in tumorigenesis and cancer progression. Indels of more than a few bases are challenging to discover with typically used alignment-based methods. In addition, most variant callers analyze tumor and normal data separately, which can introduce false positives such as when a mutation shows partial support in the normal sample. Towards addressing these shortcomings, we recently introduced Lancet, a new somatic variant caller developed under the auspices of the ITCR R21. Lancet leverages local assembly and joint analysis of tumor-normal paired data using region-focused colored de Bruijn graphs, with on- the-fly repeat composition analysis and a self-tuning k-mer strategy. This results in relatively reduced reference bias; an improved ability to detect variations that significantly diverge from the reference chromosome representations; a reduction in the scale of the analysis, leading to increased power and sensitivity to detect variants through localized, comprehensive graph exploration; and dynamic adjustment of calling behavior according to the sequence conditions of each genomic region. In testing, Lancet shows superior performance to all major alignment-based methods in terms of accuracy, particularly in the detection of ‘twilight zone’ indels (30- 250 bp). Given its continued adoption and successful application in over a dozen high-impact publications, Lancet is poised for more advanced development to enable continued improvements in its variant calling power, precision, and analytical capabilities. Specifically, Lancet is currently limited by longer runtimes than alignment- based methods, reduced sensitivity for longer insertions, lack of interactive visualization of the colored de Bruijn graph, and the inability to jointly analyze longitudinal data. To address these limitations, we propose the following Specific Aims: 1) Increase computational performance and facilitate user adoption and third-party development; 2) Add new features and enhancements to improve variant detection, phasing, and data visualization; and 3) Enable joint assembly and analysis of longitudinal data. Impact: With additional development, the next iteration of Lancet will feature advanced algorithms for fast, efficient, accurate, localized, user-friendly, and application- modifiable variant analysis of phased genome-wide timeseries data, establishing it as one of the leading methods for variant calling in cancer research.