Universal antigen encoding of T cell activation from high-dimensional cytokine dynamics

Area of Science:

• Systems immunology
• Computational biology
• Immunoinformatics

Background:

• Systems immunology needs theoretical frameworks for high-dimensional data analysis.
• Understanding CD8+ T cell activation is crucial for immunotherapy development.

Purpose of the Study:

• To develop a computational framework for analyzing high-dimensional T cell activation data.
• To model T cell responses in an antigen-specific manner using machine learning.

Main Methods:

• Combined a robotic platform with machine learning for experimental measurement and theoretical modeling.
• Utilized dimensionality reduction to create a 'latent space' for cytokine dynamics.
• Developed an 'antigen encoding' method to capture antigen-specific T cell activation patterns.

Main Results:

• High-dimensional cytokine data was compressed into a low-dimensional, antigen-specific latent space.
• The model successfully reconstructed T cell immune activation patterns.
• Identified six distinct classes of antigens based on T cell response profiles.
• Generalized the antigen encoding approach to drug perturbations and CAR T cell activation.

Conclusions:

• Antigen encoding provides a universal framework for modeling T cell activation across diverse immune settings.
• This approach facilitates the rational design and optimization of immunotherapies.
• Enables deeper theoretical understanding and predictive modeling of immune responses.