The COVIDome Explorer researcher portal

Abstract

COVID-19 pathology involves dysregulation of diverse molecular, cellular, and physiological processes. To expedite integrated and collaborative COVID-19 research, we completed multi-omics analysis of hospitalized COVID-19 patients, including matched analysis of the whole-blood transcriptome, plasma proteomics with two complementary platforms, cytokine profiling, plasma and red blood cell metabolomics, deep immune cell phenotyping by mass cytometry, and clinical data annotation. We refer to this multidimensional dataset as the COVIDome. We then created the COVIDome Explorer, an online researcher portal where the data can be analyzed and visualized in real time. We illustrate herein the use of the COVIDome dataset through a multi-omics analysis of biosignatures associated with C-reactive protein (CRP), an established marker of poor prognosis in COVID-19, revealing associations between CRP levels and damage-associated molecular patterns, depletion of protective serpins, and mitochondrial metabolism dysregulation. We expect that the COVIDome Explorer will rapidly accelerate data sharing, hypothesis testing, and discoveries worldwide.

Keywords: COVID-19; CRP; SARS; data portal; immune system; infection; inflammation; metabolism; multi-omics; serpins.

Graphical abstract

Figure 1

The COVIDome dataset (A) Schematic of experimental approach. Blood samples were collected and processed for multi-omics analysis. Created with graphic elements from BioRender.com. (B–G) (Left) Volcano plot indicating the impact of COVID-19, and (right) sina plots with boxes indicating median and interquartile range of representative features for (B) whole blood transcriptome, (C) plasma SOMAscan proteomics, (D) plasma mass spectrometry (MS) proteomics, (E) plasma cytokine profiling, (F) red blood cell MS metabolomics, and (G) mass cytometry of peripheral blood mononuclear cells (PBMCs). In the volcano plots, the vertical dashed midlines indicate no change in COVID-19 patients versus controls, and the horizontal dashed lines indicate the statistical cutoff of q < 0.1 (false discovery rate of 10% [FDR10]). The numbers at the top left and right of each volcano indicate the number of features passing the statistical cutoff. In the sina plots, q values were calculated with DESeq2 (transcriptome, adjusted for age and sex) or linear models adjusting for age and sex (all other datasets). Sample sizes range from 30 to 31 for COVID-19-negative controls and from 65 to 71 for COVID-19-positive patients, depending on the platform.

Figure 2

The COVIDome Explorer researcher portal Schematic illustrating the design of the COVIDome Explorer researcher portal and its various functionalities. Created with graphic elements from BioRender.com.

Figure 3

Cross-omics correlations enabled by the COVIDome Explorer (A, C, and E) (Top) Volcano plots for Spearman correlations between (A) MX1 mRNA levels and SOMAscan proteomics, (C) kynurenine levels from red blood cell (RBC) metabolomics and SOMAscan proteomics, and (E) plasmablast frequency and transcriptomics. The horizontal dashed lines indicated the statistical cutoff of q < 0.1 (FDR10). Numbers in the left and right quadrants indicate the number of features passing the statistical cutoff. (B, D, and F) (Bottom) Scatterplot for correlations of (B) MX1 mRNA levels with MX1 protein levels, (D) RBC kynurenine levels with WARS protein levels in plasma, and (F) plasmablast frequency with IGKV4-1 mRNA levels. Points are colored by density; lines represent linear model fit with 95% confidence interval. Sample sizes range from 65 to 71 depending on the platform.

Figure 4

CRP levels correlate with damage-associated molecular patterns (A) Sina plots showing values for CRP and ferritin light chain (FTL) measured by MS proteomics comparing COVID-19-negative (−) to COVID-19-positive (+) patients. Data are presented as modified sina plots with boxes indicating median and interquartile range. (B) Scatterplots displaying correlations between CRP levels versus SAA1, LBP, and IL-10. MSD, Meso Scale Discovery assay. Points are colored by density; lines represent linear model fit with 95% confidence interval. (C) Metascape pathway enrichment analysis of proteins detected by SOMAscan proteomics that are significantly and positively correlated with CRP. (D) Scatterplot displaying correlations between CRP levels and representative factors from systemic lupus erythematosus (CXCL10) and positive regulation of Th2 cytokine (IL-6) signatures. Points are colored by density as in (B); lines represent linear model fit with 95% confidence interval. (E) Heatmap displaying changes in circulating levels of proteins in the DNA Methylation signature that are significantly positively correlated with CRP levels. The left column represents Spearman rho values for correlation with CRP levels, while the right columns display median Z scores for each feature for COVID-19-negative (−) versus COVID19-positive patients (+). Z scores were calculated from the adjusted values for each SOMAmer in each sample, based on the mean and standard deviation of COVID-19-negative samples. Asterisks indicate a significant difference between COVID-19 patients and the control group. (F) (Top) Scatterplot for correlation of CRP with H2AFZ. Points are colored by density as in (B); lines represent linear model fit with 95% confidence interval. (Bottom) Sina plot for H2AFZ with boxes indicating median and interquartile range. (G) Heatmap displaying changes in circulating levels of proteins in the response to heat group as described for (E). (H) Data for HSPA1A as described for (F). q values in (F) and (H) are derived from linear models. Sample sizes range from 30 to 31 for COVID-19-negative controls and from 69 to 71 for COVID-19-positive patients, depending on the platform.

Figure 5

CRP levels correlate with depletion of protective serpins (A and B) Correlation analysis of CRP with SERPINA5 (A) and SERPINA4 (B). (Left) Scatterplot for correlation of CRP with the indicated SOMAmer reagent. Points are colored by density; lines represent linear model fit with 95% confidence interval. (Right) Sina plot for indicated SOMAmer reagent with boxes indicating median and interquartile range. (C) Heatmap displaying changes in circulating levels of complement and coagulation proteins significantly correlated with CRP levels with an absolute rho value greater than 0.3. The left column represents Spearman rho values, while the right columns display median Z scores for each feature for COVID-19-negative controls (−) versus COVID-19-positive patients (+). Z scores were calculated from the adjusted values for each SOMAmer in each sample, based on the mean and standard deviation of COVID-19-negative samples. Asterisks indicate a significant difference between COVID-19 patients and the control group. (D–G) Scatterplots and sina plots as in (A) for KLKB1, KLK13, C9, and C3, respectively. q values in each are derived from linear models. Sample sizes range from 30 to 31 for COVID-19-negative controls and from 69 to 71 for COVID-19-positive patients, depending on the platform.

Figure 6

CRP levels correlate with dysregulated mitochondrial metabolism in blood cells (A) Scatterplot displaying correlations between CRP levels and indicated metabolites. Points are colored by density; lines represent linear model fit with 95% confidence interval. (B) Histogram displaying the results of Ingenuity Pathway Analysis (IPA) of metabolic pathways for mRNAs measured in the whole blood transcriptome analysis that are significantly and negatively correlated with CRP. (C) Heatmap displaying expression changes in mRNAs in the oxidative phosphorylation (OXPHOS) IPA signature from (B). The left column represents Spearman rho values for correlations with CRP, while the right columns display median Z scores for each feature for COVID-19-negative controls (−) versus COVID-19-positive patients (+). Z scores were calculated from the adjusted RPKM (reads per kilobase transcript per million mapped reads) values for each mRNA in each sample, based on the mean and standard deviation of COVID-19-negative samples. Asterisks indicate a significant difference between COVID-19 patients and the control group. (D) (Left) Scatterplots for correlations between CRP and the indicated mRNAs. Points are colored by density as in (A); lines represent linear model fit with 95% confidence interval. (Right) Sina plots for indicated mRNAs with boxes indicating median and interquartile range. q values in each sina plot are from DESeq2. Sample size is 30 for COVID-19-negative controls and from 65 to 71 for COVID-19-positive patients, depending on the platform. (E) Summary of findings indicating dysregulation of mitochondrial metabolism in the bloodstream of COVID-19-positive patients with CRP levels. Created with graphic elements from BioRender.com.