An extensible brain knowledge base and toolset spanning modalities for multi-species data-driven cell types

项目摘要 大脑倡议细胞普查网络(BICCN)正在完成对成年小鼠的全面细胞普查 大脑，和大脑倡议细胞图谱网络(BICAN)将扩展这项工作，重点是人类和非 人类灵长类动物。有效地组织、汇总、访问和提炼这些地图集对于 最大限度地发挥它们对科学的影响。这项建议是为了开发一个可扩展的脑细胞知识库 (BCKB)接收和标准化来自Bcan开发的 多模式、多物种脑细胞图谱，并作为开放和互动社区传播该图谱 促进大脑知识发展的资源。BCKB将在此项目期间使用多个 来自BICCN的维度脑细胞类型分类，并将随着数据和知识的产生而扩展 比肯研究人员。 在目标1下，我们将创建一个自适应知识图，用于链接脑细胞信息。空间聚集 将使用共同的坐标框架来完成。灵活的基于图形的数据模型将捕获离散和 连续的细胞类型关系。这项工作将从跨物种拖把数据和全面的整体数据开始。 来自BICCN的小鼠脑数据，后来扩展到Bcan的全脑分子和空间 当人类和非人类灵长类动物的转录组数据可用时，这些数据就会出现。锚定 单细胞分子和空间转录组学中的分类学为整合 空间映射和/或单元类型映射的多模式数据。一个由工具组成的生态系统， 单元格类型数据的注记、映射和可视化将在AIM 2中创建。我们将构建和扩展工具， 例如，展示BICCN MOP结果的初始细胞类型卡，因此Bcan团队和公共实验室可以 分享和完善脑细胞类型分类和解剖整合。为此开发的工具框架 AIM将提供一个类似于“应用商店”的中心枢纽，以便从社区访问我们的工具和其他工具 与单元格类型数据交互。这一互联工具框架将简化科学工作流程并鼓励 公平的做法。作为目标3的一部分，我们将开发一个连接脑细胞数据和知识的基础设施。这 基础设施将能够摄取、存储、搜索和管理来自 将信息来源整合到一个相互关联的信息平台。这一知识基础设施将有助于将不同的 使用专家注释、机器学习推理和使用起源的派生的单元类型信息 机械装置。我们将使用艾伦研究所的大脑知识平台进行初步实施。最后，在目标4中 我们将通过举办年度会议，收集、管理和整合来自BICAN团队的信息和知识- 关于培训和反馈研讨会。这些活动将在Bcan项目内部和外部创建参与度 并促进基于社区的演变、可持续发展和治理。

Project Summary BRAIN Initiative Cell Census Network (BICCN) is completing a comprehensive cell census of the adult mouse brain, and BRAIN Initiative Cell Atlas Network (BICAN) will extend this work with emphasis on human and non- human primates. Effectively organizing, summarizing, accessing, and refining these atlases is critical to maximizing their impact on science. This proposal is to develop an extensible Brain Cell Knowledge Base (BCKB) to ingest and standardize comprehensive cell type information from BICAN’s development of a multimodal, multi-species brain cell atlas and disseminate that atlas as an open and interactive community resource for advancing knowledge of the brain. The BCKB will be initialized during this project with multi- dimensional brain cell type classifications from BICCN and will expand as data and knowledge are produced by BICAN researchers. Under Aim 1, we will create an adaptive knowledge graph for linking brain cell information. Spatial aggregation will be done using common coordinate frameworks. A flexible graph-based data model will capture discrete and continuous cell type relationships. The work will start with cross-species MOp data and comprehensive whole- brain mouse datasets from BICCN and later extend into BICAN’s whole-brain molecular and spatial transcriptomics data in human and non-human primates as such data becomes available. Anchoring of taxonomies in single cell molecular and spatial transcriptomics provides a robust framework for integrating multimodal data that is spatially mapped and/or cell types mapped. An ecosystem of tools for curating, annotating, mapping, and visualization of cell type data will be created in Aim 2. We will build and extend tools, such as the initial Cell Types Cards showcasing BICCN’s MOp results, so BICAN teams and public labs can share and refine brain cell type taxonomies and anatomical integration. The tools framework developed for this aim will provide a central hub, akin to an “app store,” to access our tools and others from the community to interact with cell type data. This connected tools framework will streamline scientific workflows and encourage FAIR practices. As part of Aim 3, we will develop an infrastructure to link brain cell data and knowledge. This infrastructure will enable ingesting, storing, searching, and curating neuroscientific information from multiple sources into a linked information platform. This knowledge infrastructure will help connect disparate pieces of cell type information using expert annotations, machine-learning inferences, and derivations using provenance mechanisms. We will use Allen Institute’s Brain Knowledge Platform for initial implementation. Finally, in Aim 4 we will gather, curate, and integrate information and knowledge from BICAN teams by conducting annual hands- on training and feedback workshops. These events will create engagement within and outside BICAN projects and foster community-based evolution, sustainability, and governance.