Fig. 2

Construction and analysis of a multicellular disease model (MCDM) based on scRNA-seq analysis of a mouse model of antigen-induced arthritis (AIA). a tSNE of cells (n = 16,751) pooled from all samples. b Cell type proportions (%) for individual AIA (S1 to S6) and control (H1 to H4) mice. c Heatmap of z-scores of the cluster-based means of normalised, denoised gene expression for cell type markers. d MCDM, in which cell types were represented by nodes connected by the predicted interactions between upstream regulatory genes in any cell type and their downstream target genes in any other cell type [24]. Edge width corresponds to the sum of Pearson coefficients for all interactions between two cell types, with arrows directed from upstream to downstream cell types. Edge colour corresponds to the upstream cell type. As indicated by more central positions in the MCDM, the four B-cell clusters were the most central cell types. e Bar plot depicting eigenvector cell type centrality in MCDM. Colours in all plots correspond to cell type colours in (a)