Cloud Enabled, Rigorous, Functional Assay Calibration (CERFAC)

摘要 利用基因测序鉴定高危疾病易感基因致病序列变异 开始于20世纪90年代中后期。大约在2010年，用于临床测序测试的技术从 桑格测序以靶向捕获大规模并行测序(即多基因小组测试)。这 技术的转变无疑增加了遗传易感性检测的临床实用性。但是有一个 意想不到的代价：转向多基因小组测试也增加了序列变异的速度 观察到不确定意义(VUS)。部分通过两个较旧的NCI R01(R01 CA121245和R01 CA164944)，我和我的合作者为评估方法的发展做出了重要贡献 以及这些VU的分类。继续这一轨迹，R01 CA264971的中心目标是升级严格性 和20世纪20年代胚系癌症基因变异分类的效率，正如标题中所述，是要提高的 临床多基因组间序列变异分类的严密性和有效性 癌症易感基因检测。该研究有四个目标：(1)将相关ACMG数据类型放入 更大的、逻辑上一致的集合，然后减少或消除这些集合之间的隐藏依赖；(2) 通过经验测量提高关键数据类型校准的严谨性；(3)细化量化 变异评价的贝叶斯计分系统；(4)序列变异评价的基准要素。 通过将VUS评估框架迁移到我们首创的贝叶斯积分系统，很明显 最初被归类为VU的很大一部分个别罕见的错义替换可以重新分类为 基于来自计算工具分析的一致证据，可能是良性的，也可能是致病性的 和高通量的功能检测结果。但有一个问题：计算工具和 功能分析需要严格校准，注意两者之间的独立性。流动的 从R01CA264971‘S目标1加上其全部目标2的第二条来看，我们提出的总体目标 补充内容是对使用特定云资源-Terra工作空间-的概念验证探索，以 通过以下方式更好地实现高通量/综合功能分析的严格校准，用于VUS评估 世界各地的调查团队。拟议的补编有三个目标：(1)制作一个模板 研究人员可以克隆以执行他们自己的校准的Terra工作空间；(2)在Terra工作空间内， 实施工作流程以生成用于证据校准的候选序列变体列表；以及(3) 同样在Terra工作空间内，开发可定制的Jupyter笔记本，该笔记本可以生成经验 功能测定校准，给出一组校准变量和测定结果。 拟议的项目应加强数据科学目标，即利用对现代计算的访问来 结合几种不同类型的大规模数据，加速了序列变体分类。

SUMMARY The use of gene sequencing to identify pathogenic sequence variants of high-risk disease susceptibility genes began in the mid-late 1990s. In about 2010, the technology used for clinical sequencing tests segued from Sanger sequencing to targeted capture massively parallel sequencing (i.e., multi-gene panel testing). This technological shift undoubtedly increased the clinical utility of genetic predisposition testing. But there was an unintended price: the shift to multi-gene panel testing also increased the rate at which sequence variants of uncertain significance (VUS) were observed. In part through two older NCI R01s (R01 CA121245 and R01 CA164944), my collaborators and I made important contributions to the development of methods for evaluation and classification of these VUS. Continuing that trajectory, the central goal of R01 CA264971 “Upgrading rigor and efficiency of germline cancer gene variant classification for the 2020s” is, as stated in the title, to improve both the rigor and the efficiency of classification of sequence variants observed during clinical multi-gene panel testing of cancer predisposition genes. The study has four Aims: (1) To place related ACMG data types into larger, logically consistent sets and then reduce or eliminate hidden dependencies between those sets; (2) To improve the rigor of calibration for key data types through empirical measurement; (3) To refine the quantitative Bayesian point-system for variant evaluation; and (4) To benchmark elements of sequence variant evaluation. With migration of the framework for VUS evaluation to the Bayesian points system that we pioneered, it is clear that a large fraction of individually rare missense substitutions initially classified as VUS can be reclassified to either Likely Benign or Likely Pathogenic on the basis of concordant evidence from computational tool analysis and high-throughput functional assay result. But there is a catch: both the computational tools and the functional assays need to be calibrated rigorously, with attention to independence between the two. Flowing from the second clause of R01 CA264971’s Aim 1 plus all of its Aim 2, the overall objective of our proposed Supplement is a proof-of-concept exploration of the use of a specific cloud resource – Terra workspaces – to better enable rigorous calibration of high-throughput / comprehensive functional assays for VUS evaluation by teams of investigators around the world. The proposed Supplement has three Aims: (1) To produce a template Terra workspace that investigators can clone to perform their own calibrations; (2) within the Terra workspace, To implement a workflow to generate a list of candidate sequence variants for evidence calibration; and (3) also within the Terra workspace, To develop a customizable Jupyter notebook that can generate an empirical functional assay calibration, given a set of calibration variants and the assay results. The proposed project should enhance the data science goal of leveraging access to modern computing to combine several different kinds of large-scale data, resulting in acceleration of sequence variant classification.