Data Analysis Core

组织的三维(3D)表示很容易被人脑理解，并且是 最丰富、最准确的方法来定量和全面地研究细胞状态及其 健康与疾病的关系。要生成3D组织表示，分子测量在单个 需要单元格分辨率。这些测量可以直接从完整的组织中进行，或者可替换地， 可以生成、测量组织的连续切片并将其组装成3D对象。同样重要的是 数据生成是强大的计算工具，支持首先将各种数据类型与 多尺度3D组织体积的不同分辨率；第二，识别单个细胞；第三，推导出 来自这样的3D单细胞组织模型的元特征。所使用和开发的计算工具 数据分析核心将不仅使这种淋巴组织的分析成为可能，而且还将普遍 适用于广泛的分子数据类型和组织。具体地说，数据分析核心将 提供存储数据和元数据的基础架构和计算工具，以集成不同的 将测量模式转换为多尺度图像，以生成组织的3D体素表示，以识别 对单个细胞进行3D表示，并确定细胞类型、其邻域等特征。所有的 这些分析将建立在一个开放源码的计算流水线(PorcCAT)上，该流水线是在 该实验室正在成为高度多路传输的2D和3D组织的标准分析管道 各种类型的数据。数据分析核心将使用OME-TIFF作为所有数据和 元数据。数据将存储在灵活的数据库(OpenBIS)中，该数据库可以直接交换原始数据 数据、任何处理步骤的数据以及处理流水线本身到蜂窝。数据的结构和 元数据存储可以很容易地与蜂窝的需求相协调。考虑到所有的分子测量 在我们提出的提供单元格解析信息的方案中，我们将把单元格作为一个“桶”来集成 不同的成像方式。光学显微镜方法和成像质量产生的图像 细胞学将被分割，并使用细胞标记和细胞和组织特征，不同的数据 将整合医疗设备以生成多尺度、多参数的图像。多尺度、连续的2D组织 地图将被注册以建立3D体素组织模型。将生成一个单元格解析模型 使用3D分割方法。然后将使用许多算法来派生元特征，例如 细胞形状、细胞邻域的模式、与形态特征的距离以及组织图案。这些 元特征可以在3D模型上可视化，以支持生物现象的研究。建议数 计算流水线，以及在此范围内产生的前所未有的淋巴器官数据集 该项目将提供最高质量和全面的淋巴器官3D图谱和可扩展的 可以很容易地用于其他数据类型和组织的数据处理和可视化蓝图。

Three-dimensional (3D) representations of tissues can be readily understood by the human brain and are the most informative and accurate way to quantitatively and comprehensively study cellular state and its relationships in health and disease. To generate 3D tissue representations, molecular measurements at single cell resolution are needed. These measurements can be performed directly from intact tissue, or alternatively, serial sections of a tissue can be generated, measured and assembled into a 3D object. Equally important to data generation are powerful computational tools that enable first, integration of various data types with different resolutions into multiscale 3D tissue volumes; second, to identify single cells; and third, to derive meta-features from such 3D single cell tissue models. The computational tools employed and developed by the Data Analysis Core will not only enable such analyses of the lymphatic tissues, but will also be generally applicable to a wide range of molecular data types and tissues. Specifically, the Data Analysis Core will provide the infrastructure and computational tools to store the data and metadata, to integrate the different measurement modalities into multi-scale images, to generate 3D voxel representations of tissues, to identify the single cells in 3D representation, and to determine cell types, their neighborhood and other features. All of these analyses will be built on an open source computational pipeline (histoCAT), which was developed in the lab of Dr. Bodenmiller and is emerging as a standard analysis pipeline for highly multiplexed 2D and 3D tissue data of various types. The Data Analysis Core will use OME-tiff as a standard format for all data and metadata. Data will be stored in a flexible database (openBIS) that enables straightforward exchange of raw data, data at any step of processing, and the processing pipeline itself to the HIVE. The structure of data and metadata storage can be readily harmonized to the needs of the HIVE. Given that all molecular measurements in our proposal provide single cell resolved information, we will use the single cell as a “bucket” to integrate different imaging modalities. The images generated by the optical microscopy methods and by imaging mass cytometry will be segmented, and using cell labels and cellular and tissue features, the different data modalities will be integrated to generate multiscale, multiparamter images. The multiscale, serial 2D tissue maps will be registered to build the 3D voxel tissue models. A single cell resolved model will be generated using 3D segmentation approaches. Many algorithms will then be employed to derive meta-features, such as cell shapes, patterns of cellular neighborhoods, distances to morphological features, and tissue motifs. These meta-features can be visualized on the 3D model to support the study of biological phenomena. The proposed computational pipeline, together with the unprecedented datasets generated of the lymphatic organs within this project will provide highest quality and comprehensive 3D Atlas of the lymphatic organs and a scalable blueprint of data processing and visualization that can be readily employed for other data types and tissues.