Computational tools for uniform processing and integration of human reference atlas data [2 of 5]

HuBMAP计算工具组件负责开发标准化的 所有HuBMAP数据的处理流水线，这是实现联合存储、索引、 HuBMAP数据集的显示、查询和映射。我们所倡导的一个关键创新方面 而且，据我们所知，这对于拥有多个集团、平台和数据的大型财团来说是独一无二的 实际上，我们实现的所有HuBMAP管道都提供了对所有数据的统一处理 来自相同的模式。具体地说，对于这些模式中的每一种，我们使用相同的计算 初始处理和后续处理的算法。例如，所有scRNA-Seq数据集，无论 它们是由7个平台中的哪些生成的，由我们的鲑鱼量化管道处理。 这使得所有HuBMAP数据之间的集成和比较变得容易得多。此外，所有 下游分析包括降维、聚类、差异表达等 对所有平台使用相同的方法执行，从而轻松组织数据的显示 在门户上，并为量化查询编制数据索引。类似的管道已经被开发出来用于 其他医疗模式。 在HuBMAP联盟的生产阶段，我们将把这些计算管道扩展到 支持来自新组织地图绘制中心和其他数据提供商的其他数据类型。我们会 此外，我们还将重点扩展到单细胞图像分割之外，以支持 人类参考地图集(HRA)协作，产生从细胞到功能的分层链接 从组织单位到更高层次的解剖结构。

The HuBMAP Computational Tools Component is responsible for development of standardized processing pipeline for all HuBMAP data, which is the foundation that enables joint storage, indexing, display, querying and mapping of HuBMAP datasets. A key innovative aspect that we have championed and, which as far as we know is unique for large consortia with several groups, platforms and data modalities is the fact that all HuBMAP pipelines we implemented provide uniform processing of all data from the same modality. Specifically, for each of these modalities we use the same computational algorithm for initial and downstream processing. For example, all scRNA-Seq datasets, regardless of which of the 7 platforms they were generated by, are processed by our Salmon quantification pipeline. This makes integration and comparison much easier across all HuBMAP data. In addition, all downstream analysis including dimensionality reduction, clustering, differential expression etc. are also performed using the same methods for all platforms making it easy to organize the display of the data on the portal and to index the data for quantitative queries. Similar pipelines have been developed for other modalities. In the production phase of the HuBMAP Consortium, we will extend these computational pipelines to support additional data types from new tissue mapping centers and other data providers. We will additionally extend our focus beyond single-cell image segmentation to support the mapping efforts of the Human Reference Atlas (HRA) collaboration, producing hierarchical links from cells to functional tissue units to higher-level anatomical structures.