#3120 PREDICTION OF COVID-19 SEVERITY IN HOSPITALIZED PATIENTS USING URINARY PROTEOMICS AND MACHINE LEARNING

Abstract Background and Aims COVID-19 has been a significant public health concern for the last three years; however, not much is known about mechanisms that lead to severe kidney outcomes in patients hospitalized with COVID-19. In this multicenter study, we combine isobaric TMT-tagged urinary proteomics machine learning predict patients. Method Urine samples from two medical centers (Mount Sinai Hospital University of Michigan) were used study adherence proper consenting protocols. prepared LC-MS/MS analysis as previously reported [1]. The obtained spectra analyzed using Proteome discoverer software matched against Uniprot human database. For constructing ML algorithm, randomly divided into discovery validation set at 2:1 ratio. Severe defined ICU admission, mechanical ventilation, acute injury (AKI), death, or length stay more than 21 days. Limma test was on identify differentially expressed proteins then features selected Boruta feature selection method. 10-fold cross random forest model applied obtain receiver operating characteristic (ROC) curves. Results 120 PCR-positive different collected within one week hospitalization. More 3,000 unique identified mass spectrometry. predictive stratified mild outcomes. Using set, (DEPs) outcome cohort vs (Figure 1A). A 12 top method construction 1B). generated ROC curves show algorithm demonstrated good power both 87% 79% accuracy, respectively close 90% specificity 1C, D). On average, major adverse events observed 5-13 days after Enrichment DEP compared healthy showed upregulation immune related processes downregulation proteolytic metabolic processes. DEPs exocytosis some cell adhesion extracellular matrix organization 2A, B). Upregulated associated proximal tubular cells addition pulmonary alveolar 2C). Downregulated strongly such podocytes mesangial endothelial 2D). Conclusion Here, developed an prediction severity We further delineate potential drive Learnings can be developing therapeutic options long COVID, better preparedness event other respiratory illnesses future.