Classification of COVID-19 Patients into Clinically Relevant Subsets by a Novel Machine Learning Pipeline Using...

Area of Science:

• Immunology
• Computational Biology
• Genomics

Background:

• The COVID-19 pandemic highlights the need for predictive tools to manage disease heterogeneity.
• Identifying drivers of immune pathology is crucial for stratifying patients by disease severity.

Purpose of the Study:

• To develop a machine learning pipeline for stratifying COVID-19 patients based on disease severity using blood transcriptome data.
• To differentiate severe COVID-19 cases from other forms of acute hypoxic respiratory failure.
• To identify potential blood-based biomarkers for COVID-19 diagnosis and severity.

Main Methods:

• Utilized an iterative machine learning pipeline.
• Analyzed gene enrichment profiles from blood transcriptome data.
• Compared gene expression patterns between mild/moderate and severe COVID-19 patients, and with non-COVID hypoxic respiratory failure.

Main Results:

• Gene enrichment in COVID-19 patients indicated broad cellular expansion and metabolic dysfunction.
• Severe COVID-19 cases showed specific enrichment in neutrophils, activated B cells, T-cell lymphopenia, and proinflammatory cytokines.
• Identified small gene signatures in blood predictive of COVID-19 diagnosis and severity.

Conclusions:

• The developed machine learning pipeline effectively stratifies COVID-19 patients by disease severity.
• Specific immune cell and cytokine profiles are associated with severe COVID-19.
• Identified gene signatures hold promise as clinical biomarkers for COVID-19 diagnosis and severity prediction.