Fig. 1. Host Transcriptional Signatures of SARS-CoV-2 Infection as Compared to Other Respiratory Viruses.

a Hierarchical clustering of 121 genes comprising the union of the top 50 differentially expressed (DE) genes by significance in each of the pairwise comparisons between patients with COVID-19 (SARS-CoV-2; n = 93), other viral ARIs (n = 41) and non-viral ARIs (n = 100). Gene expression values were clustered after variance-stabilizing transformation and  row normalization. Group labels and viral load of SARS-CoV-2 are shown in the annotation bars. rpM, reads-per-million. b Normalized enrichment scores of selected REACTOME pathways that achieved statistical significance in at least one of the gene set enrichment analyses, using either DE genes between SARS-CoV-2 and non-viral ARIs or between other viral ARIs and non-viral ARIs. If a pathway could not be tested in one of the comparisons since it had <10 members in the input gene set, the enrichment score was set to 0. Pathway p-values were calculated using an adaptive, multilevel splitting Monte Carlo approach and Benjamini–Hochberg adjusted. c In silico estimation of cell-type proportions in the bulk RNA sequencing using single-cell signatures. Black lines denote the median. The y-axis in each panel was trimmed at the maximum value among the three patient groups of 1.5*IQR above the third quartile, where IQR is the inter-quartile range. Pairwise comparisons between patient groups were performed with a two-sided Mann–Whitney–Wilcoxon test followed by Bonferroni’s correction. Sample sizes as in (a). d Scatter plots of normalized gene counts (log₂ scale, y-axis) as a function of SARS-CoV-2 viral load (log₁₀(rpM), x-axis). Shown are inflammasome-related genes selected from among the genes most depressed in expression in SARS-CoV-2 compared to other viral ARIs. Robust regression was performed on SARS-CoV-2 positive patients with log₁₀(rpM) ≥ 0 (n = 82) to characterize the relationship to viral load. Shaded bands represent 95% confidence intervals. Statistical results listed for each gene refer to, from top to bottom: the regression analysis (p-values for difference of the slope from 0 derived from a t-statistic and Benjamini–Hochberg adjusted; R² is the adjusted robust coefficient of determination), the DE analysis between SARS-CoV-2 and non-viral ARIs (p-values derived from a moderated t-statistic and Benjamini–Hochberg adjusted), and the DE analysis between SARS-CoV-2 and other viral ARIs (p-values derived from a moderated t-statistic and Benjamini–Hochberg adjusted). Sample sizes for DE analyses as in (a). FC, fold-change.