Identifying SARS-CoV-2 infected cells with scVDN.

Area of Science:

• Virology
• Computational Biology
• Genomics

Background:

• Single-cell RNA sequencing (scRNA-seq) is vital for studying cellular diversity in viral research.
• Identifying SARS-CoV-2 infected cells at the single-cell level is currently difficult, impeding research on viral pathogenesis and treatment development.

Purpose of the Study:

• To develop a deep learning framework for predicting single-cell infection status.
• To enable precise identification of virus-infected cells for improved diagnostics and therapeutics.

Main Methods:

• Developed the single-cell virus detection network (scVDN), a deep learning model.
• Trained scVDN on scRNA-seq data from multiple nasal swab samples with diverse cell types.
• Established a model evaluation framework for real-world experimental data.

Main Results:

• scVDN demonstrated superior performance compared to four state-of-the-art machine learning models.
• Achieved a perfect Area Under the Curve (AUC) score of 1 in four cell types, even with imbalanced datasets.
• Successfully identified SARS-CoV-2 infected cells with high accuracy.

Conclusions:

• scVDN offers a powerful tool for advancing virus research and public health.
• The framework facilitates single-cell level identification of infected cells, crucial for diagnosis and treatment.
• The scVDN model and datasets are publicly available for broader application in viral studies.