Data Analysis Unit

摘要 可重复和稳健的计算方法，再加上严格的统计 分析，是CPTCA成功的关键。我们将确保以统一的方式处理 利用现有基础架构的数据分析、集成和管理 以及我们调查人员的计算能力。我们的数据分析工作将是 分为四个层，随着我们向上移动，数据集成级别不断提高。 第一级分析中使用的方法大多是开源的，由其他 小组，而第二至第四层分析中使用的大多数方法将由 我们的团队。具体而言，纵向多组学数据的集成和可视化 带来了许多新的挑战。数据分析股(DAU)成员拥有广泛的 有统计、算法开发、基因组数据分析、大规模数据方面的经验 管理和协调多项目中心内的数据分析工作。我们 提出以下五项具体目标，以实现DAU的目标： 1)设计和实施用于一级分析的管道，使用公共和In- 内部软件工具。管道将处理使用所有分析类型生成的原始数据 CPTCA；2)开发和部署二级分析的计算方法。 这些方法将用于发现不同细胞类型并对其进行分类 肿瘤、恶性细胞克隆进化的推断和空间推断 肿瘤中细胞和基因表达模式的分布；3)发展网络- 基于第三级分析的方法。这些方法将用于发现 有助于肿瘤的时空异质性的途径；4)构建 综合肿瘤图谱。我们将汇总临床、基因组和成像数据，并 在整个项目中收集的元数据；5)与数据协调部门协作 人类肿瘤图集网络(HTAN)的其他研究中心。 与DCC和其他研究中心的调查人员合作，我们将为 对Htan调查人员生成的软件进行基准测试，开发通用 数据格式，以及提高软件工具的互操作性。

Abstract Reproducible and robust computational methods, coupled with rigorous statistical analyses, are critical for the success of CPTCA. We will ensure a unified approach to data analysis, integration, and management that leverages the existing infrastructure and the computational strengths of our investigators. Our data analysis effort will be divided into four tiers, with increasing level of data integration as we move up the tiers. Methods used in tier one analyses are mostly open-source and developed by other groups whereas most methods used in tiers two to four analyses will be developed by our team. Specifically, the integration and visualization of longitudinal multiomics data poses many new challenges. Members of the Data Analysis Unit (DAU) have extensive experience in statistics, algorithm development, genomic data analyses, large-scale data management, and coordination of data analytic efforts within multi-project centers. We propose the following five specific aims to achieve the goals of the DAU: 1) To design and implement a pipeline for tier-one analyses using both public and in- house software tools. The pipeline will handle raw data generated using all assay types by the CPTCA; 2) To develop and deploy computational methods for tier-two analyses. These methods will be used for the discovery and taxonomy of different cell types in a tumor, inference of clonal evolution of malignant cells, and inference of spatial distribution of cells and gene expression patterns in a tumor; 3) To develop network- based methods for tier-three analyses. These methods will be used for the discovery of pathways contributing to spatial and temporal heterogeneity of the tumor; 4) To construct integrated tumor atlases. We will aggregate clinical, genomic and imaging data and metadata collected throughout the project; 5) To collaborate with the Data Coordinating Center (DCC) and other research centers of the Human Tumor Atlas Network (HTAN). Working with investigators at the DCC and other research centers, we will contribute to benchmarking of software generated by HTAN investigators, development of common data formats, and improvement of interoperability of software tools.