Fig 2: Analysis of the binding site of PTX3 to hsa-miR-4766-5p. (A) Bioinformatics predicts the binding site of PTX3 to hsa-miR-4766-5p. (B) Double luciferase reports experimental test results. NC = No template control. *P < .05, compared with NC group.

Fig 3: External protein marker validation and PTX3 selection.a PTX3 was among the proteins most strongly associated with poor outcome among 1472 unique plasma proteins (Data provided by the MGH Emergency Department COVID-19 Cohort (Filbin, Goldberg, Hacohen) with Olink Proteomics)38. Of the 10 proteins we found to be associated with outcome in our DIA–MS data (Fig. 3c, d), PROC and F7 were the only proteins also measured in the external validation data, confirming an inverse association with mortality. The log2FC (adjusted for categorical age) was higher for PTX3: 0.8, adjusted P value = 0.00044 compared with PROC: -0.4, adjusted P value = 0.0006 and F7: -0.3, adjusted P value = 0.03. b PTX3 measurements by ELISA (KCH and GSTT samples for COVID-19-ICU cohorts in left and right panel, respectively). c ELISA measurements for RAGE, as an established marker for ARDS. d High-performance liquid chromatography (HPLC) fractionation of plasma (n = 35-time points from 13 patients, KCH). PTX3-containing high molecular weight (HMW) fraction is shaded in gray. A280 denotes the absorbance of the eluent at 280 nm. e Proteomics analysis of the HMW fraction. Significant Spearman correlations of PTX3 with neutrophil- and macrophage-related proteins. All statistical analyses are two-tailed.

Fig 4: SARS-CoV-2 mortality prediction using machine learning.a Kaplan–Meier plot for age (using the median age of 54 years). b Kaplan–Meier plot for SARS-CoV-2 RNAemia. As a single predictor, RNAemia provides the best stratification for survival. c Kaplan–Meier plot for PTX3 using the median levels of serum or plasma. d–f Kaplan–Meier plots for “RNAemia, PTX3”, “Age, RNAemia”, and “Age, PTX3” combined using support vector machine with radial basis function kernel (SVM RBF), a non-linear machine learning model. The machine learning model selected binary combinations of “Age, RNAemia” and “Age, PTX3” as the best predictors. Kaplan–Meier analysis is two-tailed. Nonsurvivors: n = 18; survivors: n = 60.

Fig 5: Analysis of plasma PTX3 and hsa-miR-4766-5p levels. (A) ELISA detects plasma PTX3 levels. (B) ROC analysis of the diagnostic value of plasma PTX3 levels for EHT. (C) qRT-PCR detection of plasma hsa-miR-4766-5p levels. (D) ROC analysis of plasma hsa-miR-4766-5p levels in the diagnosis of EHT. (E) Correlation between plasma PTX3 and hsa-miR-4766-5p levels in patients with EHT. (F) Correlation between plasma PTX3 and hsa-miR-4766-5p levels in the control group. EHT = essential hypertension.