Background
COVID-19 remains a major public health challenge, requiring the development of tools to improve diagnosis and inform therapeutic decisions. As dysregulated inflammation and coagulation responses have been implicated in the pathophysiology of COVID-19 and sepsis, a team from the Karolinska Institute in Stockholm studied the plasma proteome profiles of these diseases to delineate similarities from specific features. Using the Olink® Target 96 Immune Response, Inflammation and Organ Damage panels, they measured plasma proteins from COVID-19 patients in both the acute and convalescent phases, patients with sepsis from multiple sources, and healthy controls.
Outcome
Initial analysis identified 42 infection proteins linked to both COVID and sepsis, although cluster analysis indicated a higher proportion of markers associated with cytokine storms in the sepsis patients – e.g., levels of IL6, CXCL8, IL10, IL12, TNF, & IFNγ were lower in COVID patients compared to those with sepsis. Further data analysis also identified unique protein signatures associated with distinct microbiologic etiology and clinical endotypes. For example, comparing the differential expression profiles between patients with lung infections caused by SARS-CoV-2, influenza virus or bacteria revealed a shared response of 45 proteins, with 7, 6, and 5 unique proteins specific for the three pathogens.
Machine learning was then applied to the data to look for diagnostic markers to accurately discriminate COVID and sepsis. TRIM21, CASP8, NBN, FOXO1, PIK3AP1, PTN, and BID were chosen as the best markers for ML, resulting in multiple models with extremely high accuracy to discriminate COVID-19 & sepsis. Four of the top five models consisted of single proteins, the best being TRIM21 which had an AUC=1.00. All of the ML-derived protein models significantly out-performed all of the available clinical biomarkers. The final model consisted of four proteins, where higher levels of PTN and CSF1 predicted COVID-19, and higher levels of TRIM21 and CASP8 predicted sepsis.
These data extend the understanding of host responses underlying sepsis and COVID-19, indicating varying disease mechanisms with unique signatures. The diagnostic and severity signatures identified may be candidates for the development of personalized management of COVID-19 and sepsis.
