Circulating proteins to predict COVID-19 severity

Area of Science:

• Proteomics
• Immunology
• Disease Prediction

Background:

• Predicting COVID-19 severity remains difficult.
• Underlying biological pathways are not fully understood.

Purpose of the Study:

• To identify circulating proteins that can predict COVID-19 severity.
• To develop and validate prediction models for COVID-19 severity.

Main Methods:

• Measured 4701 circulating human protein abundances in 986 individuals across two cohorts.
• Trained prediction models using protein abundances and clinical factors.
• Validated models in a separate cohort of 569 individuals.

Main Results:

• A baseline model (age, sex) achieved 65% AUC for severe COVID-19 prediction in the test cohort.
• A model including 92 proteins improved AUC to 86% in the test cohort, demonstrating good generalizability.
• Selected proteins were enriched in cytokine pathways, but also non-immune pathways.

Conclusions:

• Circulating proteins measured early in disease progression are accurate predictors of COVID-19 severity.
• Protein-based models significantly enhance prediction accuracy compared to clinical factors alone.
• Further research is needed to integrate protein measurements into clinical practice.