Topological data analysis identifies distinct biomarker phenotypes during the ‘inflammatory’ phase of COVID-19
Abstract
OBJECTIVES
The relationships between baseline clinical phenotypes and the cytokine milieu of the peak ‘inflammatory’ phase of coronavirus 2019 (COVID-19) are not yet well understood. We used Topological Data Analysis (TDA), a dimensionality reduction technique to identify patterns of inflammation associated with COVID-19 severity and clinical characteristics.
DESIGN
Exploratory analysis from a multi-center prospective cohort study.
SETTING
Eight military hospitals across the United States between April 2020 and January 2021.
PATIENTS
Adult (≥18 years of age) SARS-CoV-2 positive inpatient and outpatient participants were enrolled with plasma samples selected from the putative ‘inflammatory’ phase of COVID-19, defined as 15-28 days post symptom onset.
INTERVENTIONS
None.
MEASUREMENTS AND MAIN RESULTS
Concentrations of 12 inflammatory protein biomarkers were measured using a broad dynamic range immunoassay. TDA identified 3 distinct inflammatory protein expression clusters. Peak severity (outpatient, hospitalized, ICU admission or death), Charlson Comorbidity Index (CCI), and body mass index (BMI) were evaluated with logistic regression for associations with each cluster. The study population (n=129, 33.3% female, median 41.3 years of age) included 77 outpatient, 31 inpatient, 16 ICU-level, and 5 fatal cases. Three distinct clusters were found that differed by peak disease severity (p <0.001), age (p <0.001), BMI (p<0.001), and CCI (p=0.001).
CONCLUSIONS
Exploratory clustering methods can stratify heterogeneous patient populations and identify distinct inflammation patterns associated with comorbid disease, obesity, and severe illness due to COVID-19.
Related articles
Related articles are currently not available for this article.