Cross-tissue integration of human cell identifies common cellular phenotypes across human diseases 2329

Description Human diseases affect various cell types within the local niche, but the multicellular responses of human diseases have not been collectively assessed across organs. Here, we categorize multicellular responses of human diseases based on a remapped human cell atlas, incorporating 12 million single-cell transcriptome profiles acquired from an automated data collection pipeline. Additional curation expands the collection up to 46 million cells, covering 160 disease conditions across 49 distinct organ identities. From the re-aligned atlas of single-cell transcriptome data, we constructed a hierarchical cell classification framework comprising 32 major cell types and 165 cell subtype annotations. The universal cell annotation framework enabled abundance-based clustering of the blood samples from various disease contexts, which rendered 10 different blood phenotypes including a Covid-19-specific interferon response enriched phenotype. Further, systematic profiling of disease and organ association patterns of cellular subtype reveals archetypes of human diseases repeatedly found across organs, including a multicellular phenotype highly specific to the tumor microenvironment. Collectively, our study provides foundational resources for cross-tissue cell type annotation and presents a comprehensive multicellular overview of disease-associated responses across human tissues.