Extended Data Fig. 1: Cell type annotation of PBMC, differential abundance analysis and IFN responses.

(a) Violin plots showing the expression distribution of selected canonical cell markers in the 13 clusters. The rows represent selected marker genes and the columns represent clusters. (b) Tile plot showing percentage concordance between the manually annotated 25 clusters and Azimuth annotation. (c) A bar plot of the proportion of cell types shown in Fig. 1b, separated by three conditions (n = 75 healthy, n = 9 moderate, n = 64 severe). (d) Graph representation of neighborhoods identified by Milo in COVID-19 patients. Nodes are neighborhoods, colored by their log₂ fold change between moderate (n = 8) and severe (n = 64) COVID-19 patients adjusted by age, sex, time since symptom onset and duration of systemic steroids treatment. Nhood, neighborhood. (e) Box plot showing the distribution of adjusted log₂ fold change in abundance between moderate and severe COVID-19 in neighborhoods according to 13 cell types. Boxes denote the interquartile range (IQR), and the median is shown as horizontal bars. Whiskers extend to 1.5 times the IQR, and outliers are shown as individual points. (f,g) UMAP embedding of PBMCs colored by Type I IFN response score and IFN-γ response score. The score was calculated using a gene set termed ‘GOBP_RESPONSE_TO_TYPE_I_INTERFERON’ (GO:0034340) and ‘GOBP_RESPONSE_TO_INTERFERON_GAMMA’ (GO:0034341), respectively.