A Deep Representation Learning Method for Quantitative Immune Defense Function Evaluation and Its Clinical...

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• Quantitative assessment of immune defense against pathogens is crucial for health evaluation but lacks standardized methods.
• Existing methods struggle to precisely measure the body's ability to fight off infections.

Purpose of the Study:

• To introduce ImmuDef, a novel algorithm for quantitative assessment of anti-infection immune defense function using RNA-seq data.
• To develop a reliable metric for evaluating immune status across different health conditions and diseases.

Main Methods:

• ImmuDef utilizes RNA-seq data, identifying immune signatures by comparing disease states (e.g., acquired immunodeficiency syndrome, severe sepsis) with healthy controls.
• A variational autoencoder (VAE) model (QImmuDef-VAE) reduces data dimensionality to create a latent space representation.
• A defense immune score (DImmuScore) is calculated based on the distance between patient and healthy control data points within this latent space.

Main Results:

• ImmuDef was validated on 3202 samples across four immune states, achieving high classification accuracy (71.75%-76.25%) for various infections.
• The DImmuScore effectively quantifies infectious disease severity and demonstrates strong prognostic capability, stratifying mortality/survival in sepsis and COVID-19 patients.
• The framework established a quantitative standard for cross-disease immune defense assessment, validated across five infectious diseases.

Conclusions:

• ImmuDef provides a precise, quantitative method for assessing anti-infection immune defense function using RNA-seq data.
• The DImmuScore serves as a valuable tool for evaluating immune status, predicting disease severity, and stratifying patient outcomes.
• This approach establishes a novel, standardized framework for quantitative immune defense evaluation across multiple infectious diseases.