The study developed a multi-omics approach to predict short-term COVID-19 progression in ICU patients. Analyzing data from 32 SARS-CoV-2-infected patients, including 124 clinical parameters, 271 proteins, and 782 metabolites/lipids, it identified CCL7, CA14 proteins, and hexosylceramide 18:2 as key markers. A machine learning model accurately forecasted worsening conditions up to five days in advance (79% accuracy three days before, 84% four to five days prior), complementing clinicians’ predictions. This workflow demonstrates omics-based biomarkers’ potential for ICU decision support during COVID-19 and future pandemics, offering a strategy adaptable for small cohort studies.
