A cellular map of human body based on single-cell deconvolution

AbstractA high-resolution map of human tissues and cancers is pivotal to understanding the human immune system and disentangling tumor immune evasion mechanisms, but is still missing now. Here, we apply a single-cell deconvolution strategy and construct a high-resolution cell map of human body based on the big data of Tabula Sapiens, CCLE, GTEx and TCGA across 54 healthy tissue sites and 33 cancer types. The map reveals organ-specific distributions of >1,000 immune cell types/states and ~100 cancerous cell states, and thus enables in silico charting the body distribution of various cell types/states in a user-defined manner akin to cell sorting. Through comparisons of 11,057 tumor-derived samples with 17,382 healthy samples, the map comprehensively depicts cell type-specific immunoediting patterns for different cancers. Additionally, we identified immunotherapy-related immune cell types and genes by in silico cell sorting. The map is also feasible to systematically interrogation of the organ-specific impacts of age and sex on cellular compositions. This map provides an important foundation dataset and deepens our understanding of human immune system from a holistic perspective. Here, we provide an in-house web program bulit by Django for visualizing the the distribution of any intersested cell states that express user-defined gene sets across various healthy or primary tumor tissues as well as other correlation analysis. We also provide windows and linux versions of this web tool for all users.

摘要
高分辨率的人体组织与癌症图谱对于理解人类免疫系统、阐明肿瘤免疫逃逸机制至关重要，但目前此类图谱仍付阙如。本研究采用单细胞反卷积（single-cell deconvolution）策略，基于涵盖54个健康组织位点与33种癌症类型的Tabula Sapiens、CCLE、GTEx及TCGA多组学大数据，构建了一幅高分辨率的人体细胞图谱。该图谱揭示了超过1000种免疫细胞类型/状态以及约100种癌细胞状态的器官特异性分布特征，从而支持以用户自定义的方式（类似细胞分选）开展计算机模拟（in silico）的人体各类细胞类型/状态的全身分布制图。通过比对11057份肿瘤来源样本与17382份健康样本，该图谱全面刻画了不同癌症的细胞类型特异性免疫编辑模式。此外，本研究通过计算机模拟细胞分选，鉴定出与免疫治疗相关的免疫细胞类型及基因。该图谱还可用于系统性解析年龄与性别对细胞组成的器官特异性影响。本图谱提供了重要的基础数据集，并从整体视角深化了我们对人类免疫系统的认知。本研究提供了基于Django搭建的自研网页程序，可实现对任意感兴趣的、表达用户自定义基因集的细胞状态在各类健康组织或原发性肿瘤组织中的分布进行可视化，同时支持其他相关分析。我们还为所有用户提供了该网页工具的Windows与Linux版本。