A multiomics recovery factor predicts long COVID in the IMPACC study

Area of Science:

• * Immunology and Computational Biology: Integration of multiomics data with machine learning for disease biomarker discovery.
• * Infectious Diseases: Investigating the long-term biological consequences of SARS-CoV-2 infection.

Background:

• * Long COVID (LC) affects 10%-35% of COVID-19 patients, causing persistent debilitating symptoms.
• * Understanding the biological basis of LC is crucial for developing effective treatments.

Purpose of the Study:

• * To identify a multiomics
• recovery factor
• using machine learning to predict long COVID.
• * To associate this factor with clinical outcomes and biological pathways.

Main Methods:

• * Machine learning applied to longitudinal multiomics data (PBMC transcriptomics, proteomics, metabolomics, CyTOF) from over 500 COVID-19 patients.
• * Recovery factor trained on patient-reported physical function scores.
• * Association analysis with LC status, disease severity, and immune cell frequencies; pathway enrichment analysis.

Main Results:

• * Lower recovery factor scores were observed in participants with LC compared to recovered individuals.
• * The recovery factor predicted LC risk as early as hospital admission.
• * Biomarkers for LC included increased inflammatory mediators, elevated heme metabolism, and decreased androgens, linked to reduced lymphocytes and increased myeloid cells.

Conclusions:

• * The multiomics recovery factor effectively identifies individuals at risk for long COVID.
• * This approach reveals novel biomarkers and potential therapeutic targets for long COVID.
• * Findings suggest persistent inflammation driving anemia and stress erythropoiesis underlie long COVID pathology.