Fig. 5: Detection of cell states associated with COVID-19 in a case-control cohort with a healthy atlas.

a, Overview of composition of disease (48,083 cells), control (14,426 cells) and atlas dataset (513,565 cells). b, UMAP embedding of cells from the COVID-19 and healthy datasets integrated with a CR (joint embedding, top) or ACR (bottom) design. Cells are colored according to disease condition (left), broad annotated cell type (middle) and expression of IFN signature (right). Mono, monocyte; prolif., proliferative; RBC, red blood cell; T_reg, regulatory T. c, Scatterplot of neighborhood DA log fold change against the mean expression of IFN signature with the ACR (left) and CR (right) designs. Neighborhoods where enrichment in COVID-19 cells was significant (log fold change > 0 and 10% spatial FDR) are colored. Pearson correlation coefficients and P values for the significance of the correlation are reported (two-sided test). d, Precision–recall curves for the detection of IFN-activated neighborhoods with DA log fold change for alternative designs (ACR or CR) and using joint embedding of reference and disease datasets (scVI) or transfer learning (scArches scVI). The AUPRC is reported in the legend, with the 95% CI calculated from bootstrapping with 1,000 resamplings shown in brackets. The dashed lines denote the baseline value for the AUPRC, indicating the case of a random classifier. e, Scatterplot of neighborhood DA log fold change against the mean expression of IFN signature with the ACR design for neighborhoods of CD14⁺ monocytes. The colored points indicate neighborhoods where the enrichment in COVID-19 cells was significant (10% spatial FDR). Neighborhoods are colored according to IFN phenotype. f, Distribution of IFN signature score for cells belonging to neighborhoods assigned to three alternative CD14⁺ phenotypes. g, Distribution of COVID-19-enriched CD14⁺ phenotypes across patients with varying disease severity (healthy: n = 23 patients; asymptomatic: n = 9 patients; mild: n = 23 patients; moderate: n = 30 patients; critical: n = 15 patients; severe: n = 13 patients). Each point represents a donor; the y axis shows the fraction of all CD14⁺ monocytes in that donor showing an IFN^hi COVID-19-enriched phenotype (orange) and an IFN^lo COVID-19-enriched phenotype (yellow). The remaining fraction represents monocytes with a healthy phenotype (not shown). In the box plots, the center line denotes the median; the box limits denote the first and third quartiles; and the whiskers denote 1.5× the IQR.