Characterizing Macrophages Diversity in COVID-19 Patients Using Deep Learning.

Area of Science:

• Immunology
• Computational Biology
• Genomics

Background:

• Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) causes COVID-19 with variable patient outcomes.
• Understanding molecular differences in immune responses is crucial for managing COVID-19 severity.

Purpose of the Study:

• To identify molecular states associated with COVID-19 infection severity using single-cell RNA sequencing data.
• To analyze cell-type composition and inflammatory responses in mild, severe, and uninfected individuals.

Main Methods:

• Computationally processed single-cell RNA-Seq (scRNA-Seq) data from 12 Bronchoalveolar Lavage Fluid (BALF) samples.
• Analyzed cell-type composition, focusing on macrophages and T cells.
• Developed and validated artificial neural network (ANN) and graph convolutional neural network (GCNN) models for infection prediction.

Main Results:

• Significant differences in cell-type composition were observed between mild, severe, and normal infection groups.
• Macrophage populations increased significantly with infection severity (10.56% in normal to 34.15% in severe).
• T cell populations decreased significantly in severe infections (24.76% in mild to 7.35% in severe).
• GCNN models achieved 91.16% prediction accuracy for infection using macrophage subtype data.

Conclusions:

• Distinct gene expression profiles in inflammatory and immune cells characterize severe COVID-19.
• Macrophage and T cell dynamics are key indicators of COVID-19 severity.
• Computational models, particularly GCNNs, show promise in predicting COVID-19 infection severity.